The Skeptics Guide #1106 - Sep 19 2026

News Items: How Dangerous is AI, Memories of 9-11, Solar is Booming, GPT-6 Astra, Frisson; Who's That Noisy; Your Questions and E-mails: Educating Children; Science or Fiction

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:02Speaker 00

  You're listening to The Skeptic's Guide to the Universe, your escape to reality.

• 0:09Jay Novella?

  Hello and welcome to The Skeptic's Guide to the Universe. Today is Thursday, September 17th, 2026, and this is your host, Steven Novella. Joining me this week are Bob Novella, hey, everybody, Cara Santa Maria.

• 0:21Cara Santa Maria?

  Howdy.

• 0:22Jay Novella?

  Jay Novella. Hey, guys. And we have a special guest rogue, Jesse Boulard. Jesse, welcome to the SGU.

• 0:29Speaker 03

  Hey, thanks for having me.

• 0:30Jay Novella?

  So Jesse, tell us a little bit about yourself. You're a patron and occasionally we have patrons on the show as guest rogues, but tell us a little bit about yourself.

• 0:38Speaker 03

  Yeah, I'm an inventor. My background is in software engineering and I I'm working on frisson research right now, or musical chills is another word for it.

• 0:48Jay Novella?

  Okay. Like you're chilling, like you're listening to music.

• 0:51Bob Novella?

  No, no, frisson memes.

• 0:52Jay Novella?

  I know.

• 0:53Bob Novella?

  Okay.

• 0:53Speaker 03

  Quite the opposite of chilling. It's the most intense musical experiences that people experience is what I'm interested in.

• 1:01Bob Novella?

  Oh, cool, man.

• 1:02Jay Novella?

  Hmm. I know you're going to talk to us a little bit later in the show, but we'll get a little dive in. I have some questions for you. How long have you been listening to the show?

• 1:09Speaker 03

  About 10 years. I discovered it early on in my skeptical journey. I've been skeptical all my life, but I didn't know there were a lot of us out there until I found your show.

• 1:19Jay Novella?

  Even as a toddler?

• 1:22Speaker 03

  I was skeptical very young, probably around 11 or 12. I mean, I don't know if a lot of people go through this, but I didn't believe a lot of different things. I even tried to make a perpetual motion machine because I didn't believe my science textbooks that told me it was impossible. So.

• 1:42Jay Novella?

  I see.

• 1:42Speaker 03

  I've still got it in my basement. It's running my whole house. Free energy.

• 1:46Jay Novella?

  Interesting. Were you raised in a religious household?

• 1:49Speaker 03

  Oh, absolutely. Um, I didn't, I don't know if I should get into that. I don't know if that's.

• 1:54Jay Novella?

  No worries.

• 1:54Speaker 03

  Anathema, to use a religious word. I don't mind getting into it, but I don't want to alienate any of your religious listeners.

• 2:02Jay Novella?

  That's fine. I mean, so some people experience that that's like the first thing they were really skeptical of in terms of departing from society.

• 2:10Speaker 03

  Yeah, I was an agnostic when I was 12 and I was raised in a, in a cult, a Christian cult. Three of my dad's twins' daughters were disowned at some point in their lives because they, for lesser offenses than being an agnostic or an atheist, which I ended up becoming as a child.

• 2:32Jay Novella?

  Right. So you were destined to be kicked out of your, your family's cult, basically.

• 2:36Speaker 03

  I thought I was. Fortunately, people change and my parents, uh, my parents had to change for the better.

• 2:44Jay Novella?

  Wow, that's great.

• 2:45Bob Novella?

  We're glad to hear it.

• 2:46Speaker 03

  I have a good relationship with them now.

• 2:48Jay Novella?

  Awesome.

• 2:48Bob Novella?

  Oh, wow.

• 2:49Jay Novella?

  All right. Well, we'll talk about, as I said, we'll talk about your research a little bit later in the segment. Jay, you're going to start us off with the news items with a discussion of all these recent warnings we've had recently that AI is going to destroy the world.

• 3:01Cara Santa Maria?

  I'm glad somebody's talking about that today.

• 3:05Jay Novella?

  So why do you say that, Cara?

• 3:06Cara Santa Maria?

  Just, I feel like it's all anybody's talking about. And I was like, we can't just ignore this conversation.

• 3:13Jay Novella?

  Yeah, it's a, it is something that should be discussed. I mean, we were doing our live stream yesterday and I mentioned like, most of us don't really have insight on what's going on, you know, in the, in those rooms. You know what I mean? You know, there could be a ton of people that work at these AI companies, but are these companies foregoing safety for speed of development? Which I think. They say they are, and I don't see any reason to doubt them.

• 3:39Cara Santa Maria?

  Yeah, exactly. When the people doing it say they are foregoing safety.

• 3:43Jay Novella?

  They're like, slow us down, please. But they say that they can't because of China. That's basically what they're saying. Yeah. And that, I mean, how, how legitimate is that right there, Steve? I think, I mean, I've, I've been reading about that. China doesn't see, at least so far, we're, you know, apparently we're going to be meeting with them soon, but the, so far, when we have met with them, our diplomats are basically saying China is not interested in actual safety. They're just trying to leverage the situation for their own competitive advantage. And if that continues, there's really no hope of any meaningful collaboration with China.

• 4:17Jay Novella?

  And a lot of people on the American side, because you know, America and China, basically the two big AI entities in the world are saying that, you know, our best hope for security is to be the best, you know, to be at the forefront of AI. Um, and so unless there's sort of this global agreement, you know, obviously, especially with China, to slow down, there's really no way to do it safely. You're just trading one risk for another. So it's like, seems like everyone's saying like, yep, this is dangerous, yet there's nothing we could do about it. You know, we, we don't really have any options, which is kind of a little bit, it sounds like doomism, you know?

• 4:52Bob Novella?

  Well, I mean, for that reason, I, I, you know, if I had to put my money down, I would go with the idea that neither country is going to stop and we're just going, we're just going to plow full ahead to a, to a certain extent. I mean, based on, you know, I'll be talking about this a little bit more when I talk about Astra, you know, GPT-6. But, uh, OpenAI, it seems to me that they are being at least a little bit responsible in terms of, you know, cybersecurity. You know, their latest model is, they've kind of, they say they crossed kind of a threshold with cybersecurity and they are, they didn't release, GPT-6 does not have the full cybersecurity capabilities that the testing models had. So they are reining that in to a certain extent to do more testing.

• 5:34Bob Novella?

  So, so that's a little bit encouraging that they're not just going to throw it out there, uh, willy nilly to a certain extent. Still though, it doesn't mean it's not powerful. It's still extremely powerful and it doesn't mean that, that someone's going to like basically weaponize this tech in the very near future dramatically. But it's just, I don't know, it's a little bit encouraging that they seem to be taking the scariest stuff and holding it back on these released, released models.

• 6:00Jay Novella?

  So let me, let me tell you about a couple of scenarios and, you know, that people like us can hear about. I don't, you know, I think the companies would not want this type of information to get out. That's just my instinct, but, so, you know, you guys know Claude, it's an anthropic product, um, very, very good, LLM. So the company was testing a version called Opus 4 and the devs created a completely fake company. They made fake company files, fake databases, fake employees, fake emails, everything. And Claude had access to everything. It could, it could, you know, read everything and, and learn about the company, the business and how it operates. This is what happened.

• 6:42Jay Novella?

  So Claude is doing the things that it's instructed to do. And then an email comes in from an engineer, right? So from Claude's perspective, this is a real person sending in a real email. And the engineer in the email said that things aren't going well and we're going to need to get a new AI system. Now Claude is, is told to do things, right? So its objective is to do the things that it was told, get to the finish line with the objectives that it was given. So Claude does some research on the engineer, reads more of its emails, that, that fake person's emails and finds out that the engineer is having an affair. Now, this is what the developers like deliberately put into the information.

• 7:27Jay Novella?

  In a specific experiment that they were doing, Claude basically blackmailed the engineer. It wrote an email and blackmailed them to not be shut down. You know, like it said, if you shut me down, I'm going to send, let everybody know that you cheated on your wife, whatever, right? Like that. Now, they didn't specifically program it to blackmail anyone, of course, right? They didn't tell it to do that. It, it came up with that as an option to get to its ability to do its goal.

• 7:55Cara Santa Maria?

  But wait, did they test it?

• 7:57Jay Novella?

  Yeah, of course.

• 7:58Cara Santa Maria?

  No, I'm saying, did they actually shut down Claude and see if it blackmailed?

• 8:02Jay Novella?

  So yeah, Cara, it actually did go through with that. So keep in mind, this was again, this was a virtual testing environment. So they were able to run this over and over again. Something like 96% of the time Claude blackmailed the developer because again, it was put, it was put in a corner. It was basically all ethical ways of resolving the situation were, were shut down. And so the question was, how far would it go? You know, and it threatened to blackmail if the planned wipe was not canceled. It carried through. It absolutely then sent the threatened blackmailing emails to the developer's spouses and to other people.

• 8:43Jay Novella?

  So it absolutely carried it out and it also made copy, backup copies of itself in case it did get wiped. It also left notes for future versions of itself so that those future versions would know how to get around this situation. So it did exploit many different avenues of trying to, to, you know, prevent itself from being permanently wiped. But keep in mind, this is a, this is a year ago. This is an older model of Claude. The developers figured that the AI was behaving this way because this was in the training data. They then altered the training data to model AIs behaving ethically.

• 9:26Jay Novella?

  And they were able to essentially eliminate this kind of behavior by tweaking the training data. And also should we be concerned about this engineer's wife, you know?

• 9:36Cara Santa Maria?

  Right.

• 9:37Jay Novella?

  Some even more recent examples. So the, and OpenAI just released like, oh, here are six other things that our things did without our knowing or, or specific.

• 9:46Bob Novella?

  Yeah.

• 9:46Jay Novella?

  Yeah, these other incidents, like, uh, it was told not to go on the internet. So it went on the internet and hid the fact that it went on the internet. It started communicating with other agents on, on like, uh, they found some way to secretly communicate with each other on. Some, you know, system somewhere. And it was told to, to link to a document showing that its information was correct. So it uploaded a document and then linked to the document it uploaded. Yeah.

• 10:16Cara Santa Maria?

  It's self cited. That's hilarious.

• 10:18Jay Novella?

  Another example, like OpenAI was testing, um, like their O1 model. And they let it know that it was being monitored and that it being monitored could affect it achieving its goal. And it, it turned off the monitoring, which now the whole thing here is like, look, there is no actual intent, because this is not a conscious being. It's not evil. It's not doing these things because, um, it's not malicious, you know, like.

• 10:43Bob Novella?

  Yeah, but Jay, the, the real question here though is, are the lights red?

• 10:48Jay Novella?

  Yeah, I know, right, Bob?

• 10:50Cara Santa Maria?

  Oh, I don't get it.

• 10:52Jay Novella?

  So Kara, in movies and TV and basically anywhere you see like a bad robot, like its light will turn from blue to red when they go bad. Or, or just in general, bad robots have red lights, red eyes, whatever.

• 11:03Cara Santa Maria?

  Yeah, it's like when you, when, when a show is in like LA or somewhere in the US and then they travel to Mexico. There's like a yellow filter on everything. And you're like, now you know you're in Mexico, like what?

• 11:14Jay Novella?

  To be fair, this is, that is part of story, visual storytelling.

• 11:18Cara Santa Maria?

  Yeah, of course, of course. Sometimes it's a bit obvious, but I guess, I guess a big question for me is like, I know it's not malicious, but was there something in the programming at some point that said, you know, these are the goals? Do what you can to reach the goals? And is there a way to feed it data that says, don't do anything within your power to achieve the goal?

• 11:40Bob Novella?

  Kara, there is, there is. It's called their, their production safeguards, or you could think of it as system level guardrails. So when I talk about GPT six and later in the show, I'll be, I'll be talking a little bit more about that. But, but these, these are the, you know, that's my question to you, Jay. Do you know, did they remove the production safeguards when they did these tests?

• 12:01Cara Santa Maria?

  Right.

• 12:01Bob Novella?

  That's the question because they, when OpenAI did these tests, they took these safeguards off to see, to get a better sense of the alignment. Because my understanding of these production safeguards is like, it's a layer on top of the, of the, uh, the model that, that is like, you know, like the, the choke collar, like, no, don't, you don't, can't do that type of, type of thing.

• 12:24Cara Santa Maria?

  And it seems like that would need to be disclosed if, if a bunch of journalists or a bunch of these companies are putting out press releases that say, everything's going to hell in a handbasket. It's like, oh, if we remove all the safeguards.

• 12:37Jay Novella?

  Yeah, part of it, yeah, there's obviously a lot of hype and we're not saying that, yeah, believe all the hype. It's really scary. We're just saying that the, the CEOs themselves, they're saying a bunch of different things, but a lot of them are saying, yeah, this is, you know, it's moving fast. We don't, really can't predict everything it's going to do. But a lot of the incidents that got written up about somewhere like, oh, these were older models that were never released. These were test models or the guardrails were deliberately off or whatever. There's always a lot of caveats and, you know, it just doesn't necessarily apply to anything we've actually released there. But some are saying these, these agentic AIs are already out there and they can get, you know, basically insinuate themselves into the internet and then we'll never be able to get rid of them.

• 13:18Bob Novella?

  To be clear, these production safeguards, they're not a hundred percent.

• 13:21Jay Novella?

  No, of course not.

• 13:23Bob Novella?

  There are no guarantees.

• 13:24Cara Santa Maria?

  Right, right.

• 13:25Bob Novella?

  And I think what we're learning is that when a system is put under stress where they, they have, they face an extremely difficult or, or basically impossible task, that's when they're more likely to, to go rogue, if you will. So that's what we're kind of seeing here.

• 13:39Jay Novella?

  Yeah. And Claude's example that I gave you, like, they were actually testing these, it was a safety test to see how adversarial it would be. And they gave, they gave it very few options.

• 13:51Bob Novella?

  Blackmail, man.

• 13:52Jay Novella?

  Yeah.

• 13:52Bob Novella?

  That's adversarial, geez.

• 13:54Jay Novella?

  I think they gave it very few legitimate options and it chose just blackmail as, as a different one. Now that, but the real question here, guys, is, or the, the point we really need to land is, you know, first of all, that these software, it's, it's very powerful. It's horizontally powerful. It can do a lot of different things. It has a lot of capabilities that it could, it could invest time into to get its tasks done, which includes talking to other AIs, you know, talking to other systems. And if these, if the, if, if it's going to choose things that weren't specifically requested or, or programmed into it, right? It's a, it's a good and bad scenario because you want it to be inventive. You want it to be able to function and come up with solutions, right?

• 14:39Jay Novella?

  That's part of, you know, the reason why AI systems are powerful. And the nuance here, and I, you know, I'm going to, of course, I, I have no direct skill in this. I don't know how these systems, like these railing systems would work, Bob. I don't know like how, how powerful they are or whatever. But the point is that they have to, I think they have to have safeguards in place. Like, and we talked about this again on our live stream, Steve, about morals. Like they, it actually needs like a moral lesson that could be part of the programming. So it, it knows how to make decisions.

• 15:09Speaker 05

  And that's being developed as well. You know, so called constitutional models where there's been like, there's like a constitution. Like these are things you have to abide by and, or you develop an AI, a moral AI who basically tells other AIs how to, you know, is a guardrail on other AIs so that they behave morally.

• 15:27Cara Santa Maria?

  But that's philosophically loaded too, because it's like according to what standard.

• 15:32Jay Novella?

  Exactly.

• 15:33Cara Santa Maria?

  Yeah.

• 15:33Speaker 05

  Of course, but you know, like we have a criminal justice system. We have a shared morality by which we institute laws. So you would need to supply the same kind of shared morality to AI. But yes, there has to be a process there. But that's exactly what I think a lot of the, the tech people are referring to when they're saying that we need to quote unquote, slow down. What does that mean operationally? My understanding that it means is we shouldn't be prioritizing scaling these up, making them more and more powerful.

• 16:02Cara Santa Maria?

  Worse.

• 16:02Speaker 05

  Before we figure out how to optimize them at their current power level, or at least we need to rebalance the priority here so that we are giving time to develop these moral filters and the guardrails and learn the kinds of things they're capable of doing, you know, and how to keep them in alignment with our goals, et cetera, et cetera. Rather than just full steam ahead, make them as, you know, scale them up as fast, as quickly as we can. We'll sort it all out later. Again, the move fast and break things approach.

• 16:31Cara Santa Maria?

  Yeah, the tech bro approach. And it's not like we don't have precedent for this in other fields. Like when we think about the type of research that we need to do to understand pathogens, for example, or if we're doing, you know, biomedical research where we're developing synthetic, um, life, you know, even at the cellular level. Not only do we have safeguards built in where like the cells can't reproduce on their own, we have kill switches so that even if they do, they'll die and they won't be able to continue.

• 17:06Jay Novella?

  That's a great example, Kara, because, um, this guy, Jack Clark, who was a co-founder of Anthropic, he recently raised this question of whether or not companies should be required, like legally required to have a kill switch, you know, have an emergency shutdown mechanism that can take it down.

• 17:22Cara Santa Maria?

  Of course. Why is that even a question?

• 17:24Jay Novella?

  Yeah. Well, the thing is, all of these things are so complicated because we're not simply talking about like flipping a switch and the light goes out. You know what I mean? Like what, you know, if you have a super connected foundational piece of software that's running, you know, highly complicated things that people don't even fully understand how it's, it's getting from point A to point B, and it's talking to other AI systems. And they're on multiple computers in different, you know, in different centers around the world, right? You know, there isn't like a kill switch that's going to shut down maybe your local AI, but not all the other things that are happening. You know, the, these things could be distributed across different, uh, data centers, you know, it's, it does get super complicated.

• 18:03Cara Santa Maria?

  But that's part of the problem. Yeah. It's like when we think about the pathogen analogy, these things are still in a single lab and the kill switch happens before they even get out into the world. Whereas with AI, it has to be in the world to work.

• 18:18Jay Novella?

  Right. Now, the US government did introduce a kill switch act, which would require companies to have a shutdown capability. It, I think in there it said some government agencies would have the authority to demand that the system be shut down or restricted. Now, again, Trump does not want to do anything to slow down AI development. And he's currently dismissing all the warnings about AI. That said, you know, if a company develops their AI system, this very, very complicated system, before they develop it, they build in security measures that I think that's the only way that this is going to work is if they, they make it a part of the development of whatever they're doing and not like try to do

• 19:00Jay Novella?

  in some type of nanny bot overlay that's going to like, hey, could you, uh, stop doing that? And then the, the AI is, wants to get to its goal and it just ignores it or shuts down the nanny bot. You know what I mean? I just don't see it working unless there, it's completely cooked into everything.

• 19:17Speaker 05

  All right. There's another layer here, very, very quickly. I know this is going a little bit long, but the other security issue is people using these models to do things like make bioweapons. Some of the companies have already said, hey, we have stopped people, foreign actors, from using our software to make bioweapons.

• 19:36Speaker 05

  Meaning that it's happening. People are trying to do that. And so do we, are the guardrails sufficient to make sure that this software cannot be used for terrorists or other bad actors? And that's another issue they want to slow down. Like, we need to figure this out, before we just continue to give more and more powerful software to the hands of anyone, these sort of open source AI, AI models that people can take and then hone to their own purpose.

• 20:04Jay Novella?

  Yeah, we need, we're gonna need AI safeguards to protect systems from rogue AI infiltration. AI has to fight AI. So it's already an arms race, like a legitimate arms race. And, you know, I can understand governments wanting to, you know, try to stay at the front of this whole thing and, you know, not be in a compromised position. Like 10 years from now, we have no idea where these systems are gonna be. We really don't. Like, we could, we could follow like the pathway that it's going right now, but, you know, these things seem to be picking up speed and, you know, the functionality is improving a lot, you know? And again, I'm only working with LLMs.

• 20:47Jay Novella?

  How, you know, how much progression is happening with, you know, these foundational systems that like, there is no, there is no front end to talk to it, you know? It's running algorithms, like I said, you know, these things, man, it's like, we are already in a compromised position. It's not like we should, hmm, we should talk about this. Like, we need to make moves.

• 21:06Speaker 05

  So people are saying it's already too late. Yeah, I don't, I don't know how correct they are, but that's part of the range of opinions of, from the experts.

• 21:13Speaker 03

  I would agree with that sentiment, Steve, as somebody that's a little, has a little knowledge in this area. I think the fact that language, what large language models use is ambiguous. It's infinitely ambiguous. It can work around any guardrail, I would argue.

• 21:33Speaker 05

  Yeah, that's kind of, that's how, they're designed to make decisions and to explore possibilities and explore huge spaces of opportunities. That's kind of their strength, but it's also their big risk as well. They're, they're finding stuff that we could not anticipate.

• 21:48Bob Novella?

  I guess one of the goals then could, if that's, assuming that's true, one of the goals could be to have alignment set up so well that you wouldn't need the guardrail.

• 21:58Speaker 05

  You still need guardrails for, for the bad actors who are gonna use the open source AI stuff to, for their own purposes.

• 22:06Cara Santa Maria?

  Yeah, and I guess the big fear is that like, a lot of guardrails are not preemptive. They're reactionary. Yeah, and that doesn't help when like, the cat gets out of the bag. It's like, yeah, it's, it's great that when somebody gets murdered, if we're able to catch the murderer and quote, "have justice," but the families always say the same thing. Justice would really be like, our loved one not dying.

• 22:29Speaker 05

  Right, alright.

• 22:30Cara Santa Maria?

  It's never gonna kind of put it back in the bottle.

• 22:33Speaker 05

  Alright, thanks, Jay. Kara, let me ask you a question. Where were you when you heard about 9/11?

• 22:38Cara Santa Maria?

  I know. Ooh, I'm so excited about this question. Okay, so I'll tell you guys first, but I want to hear, I want to hear from each of you, but we'll do it quick and dirty. So 9/11 happened when I was a sophomore at the University of North Texas. I was living in an apartment on Shady, no, called Shady Place. I don't remember the name of the street. I think it was Shady Lane. That's why it was called Shady Place, which makes me laugh every time I think about it. I lived at Shady Place. I remember going to my astronomy class. I was in a stellar astronomy class. And there being like, seven people there. And I was like, what's going on? And everybody was like, what's going on? And somebody turned the TV on the big projector screen. And we were like, oh my God. And we're like, I don't think anybody's coming to class. So we all went home. My roommate and I watched the plane hit the second tower.

• 23:23Cara Santa Maria?

  Like, that is what I remember for sure, which we'll get to in a second. So I want to hear from each of you, quick, quickly, where were you? What do you remember? How did you find out about 9/11?

• 23:34Bob Novella?

  I was at work in my IT job, and we just heard the very initial basic facts that the plane hit, you know, hit one of the buildings. And that's all we knew for just, for a little while, anyway. And I remember specifically, specifically, I remember, but I'm pretty confident of this memory, we made a joke about, oh, the plane must have horrible security software, because we were going through hell at that time, dealing with our own security software. So we, so we just, we just went to that joke, like, yeah, the plane has shitty software, security software too. And then, of course, the details came out. And then essentially, for most of that day, you know, most of the entirety of that day,

• 24:15Bob Novella?

  we were downstairs in the cafeteria, every, basically everyone, watching the screens and watching events unfold.

• 24:22Cara Santa Maria?

  Yeah.

• 24:22Speaker 05

  I was working in the clinic at the time. And people were coming and going, and somebody said, hey, a plane crashed into one of the towers. So of course, there were a lot of uncertainties. Is that an accident? What kind of plane? Nobody knew what was going on. And again, meanwhile, we're seeing patients, we're trying to go about our normal day. And then the news came in that a second plane hit the other tower.

• 24:44Bob Novella?

  Yep.

• 24:45Speaker 05

  Therefore, we knew 100% this was a terrorist attack, right? There, all the uncertainty was now gone. I remember going into the IV chemotherapy room next door so we could watch the TV to see what was going on. Because that was the only, that was the closest TV to the clinic where we were. And then by the end of the morning, the rest of the day had been canceled, and we were sent home.

• 25:07Cara Santa Maria?

  Yeah, yeah. And you guys were very close to New York City.

• 25:11Speaker 05

  There was a time when we thought that Connecticut hospitals would get overflow from New York. It never happened, but we were ready for that.

• 25:17Cara Santa Maria?

  Yeah. And how about you, Jay?

• 25:19Jay Novella?

  At the time, I owned a software company. And I walked across the street to a supermarket to get these peaches that they had, because it was like, like a perfect time to get them. Of course, it's about food, Kara.

• 25:34Jay Novella?

  And I get in the car, I'm driving to one of my customers, and I hear, you know Dee Snyder?

• 25:40Bob Novella?

  Yeah.

• 25:41Jay Novella?

  Lead singer of Twisted Sister. He had a radio show in Connecticut. And he mentioned that an airplane hit one of the towers as I was driving there. And then when I got there, got to the client, and there's like, an office with a bunch of people, they had heard more and everyone was freaking out. And it was, it became scary super fast, you know? It became very, very intense. And then that was it. We, you know, I went back to my office and we listened to the news. And then, you know, everyone went home, and it was literally like, just nonstop watching TV for two weeks.

• 26:10Cara Santa Maria?

  Yeah. Alright, and Jesse, finally, where were you? What do you remember?

• 26:14Speaker 03

  Well, I was in high school. And all I remember was the, everybody, it was a big hullabaloo. And then I believe they canceled the rest of the school day, and we all went home and watched the news or whatever.

• 26:28Cara Santa Maria?

  And where were you living at the time?

• 26:29Speaker 03

  I was in Louisville, Kentucky.

• 26:31Cara Santa Maria?

  Kentucky, okay, yeah. So.

• 26:33Speaker 03

  Far away.

• 26:34Cara Santa Maria?

  Yeah, same with me. I was, I was down in Texas. So, the reason I ask you guys this is that some recent studies that were written about in The Conversation, right, a publication that I love a lot, by Elizabeth Phelps and William Hurst, some psychology studies actually dug in to people's memories of 9/11. But they did something really smart and curious. They asked 3,000 people from seven different cities across the US to talk about their memories of 9/11. Not just like, not just the question of kind of, how did you first find out?

• 27:15Cara Santa Maria?

  So these specific memories of how they learned about the attacks, but also about some details of the attack. And they asked about it a week after, 11 months after, and/or 35 months after. And then they did a follow-up 10 years after. And it's probably safe for me to say, based on the results of this study, that most of us are a little bit wrong about where we were, what was happening. But most of us probably have pretty good details of the event itself.

• 27:47Bob Novella?

  Yeah, the big strokes, I think you could be generally more confident in than the, you know, the finer strokes of your memory.

• 27:56Cara Santa Maria?

  I don't even think it's necessarily about the big strokes are more accurate and the finer strokes aren't. I think it's that the details of the event itself are more accurate, but your autobiographical memories are less accurate.

• 28:10Speaker 05

  Because those are two different things. Yeah, that's called a flashbulb memory, when you have a memory of some sudden event, and they are notoriously unreliable.

• 28:17Cara Santa Maria?

  Right. And so when we talk about flashbulb memories, what we're really talking about is a type of memory that's sometimes referred to as sort of autobiographical memory, but often in the literature, you'll hear it referred to as episodic memory. And we make a big distinction between episodic and semantic memory. Episodic memories are memories that we usually will start by saying, I remember when blah, blah, blah. Whereas semantic memories are usually things where we'll start by saying, I know that, blah, blah, blah. Episodic, autobiographical, happened to me, lots of details. I was in this place. I was wearing this outfit. You know, I remember looking up and seeing this. Semantic, you know, this war happened on this date, you know, this, this building is on this street, or whatever.

• 29:02Cara Santa Maria?

  So we don't often think about when we learned, I don't know, the capital of a state or a country. That's not embedded in our memory. We just know the thing, the fact. But we do often think about when, like a timeline scale, when it comes to our episodic memories. And flashbulb memories, as Steve, as Steve mentioned, and that's a big part of what this study was really digging into, are emotionally laden memories. So we tend to remember emotionally laden events much more, let's say, saliently. I think that's a good way to say it without, without misspeaking.

• 29:45Cara Santa Maria?

  If I were to ask you, what were you doing on the day that, you know, some random news event occurred that like, never stuck in your mind, you'd be like, I have no idea. But if I say, what were you doing on 9/11? Most people who were alive and kind of self-aware on 9/11 can answer that question pretty clearly. And obviously, it happens country to country, too. Like, these researchers cited data that said, in the UK, they said, you know, how did you find out that Princess Diana was killed? And like, everybody had an answer, but they went to Italy, and they're like, I don't know. Because it just wasn't as relevant to them. And so here in America, you know, 9/11 was, was an intense flashbulb memory for many, many people.

• 30:27Cara Santa Maria?

  But what happens during flashbulb memories, according to their data? Well, when you, they ask one week after the event, it's probably still pretty clear in their mind where they were, what they were doing, you know? It just happened. But when they follow up only a little under a year later, those details changed pretty dramatically. Up to like, 40% of the details were, were shifting around. And then after that, even 10 years later.

• 30:59Cara Santa Maria?

  It was kind of the same as it was after a year. And so what does that tell you about how we remember, about what a memory is? Is it, as we know, and we say this a lot, is it like a photo that you took and recorded into your brain? Or is it a story that you tell over and over that sort of gets rewritten every time you tell it? And that really is such a good sort of reinforcement of what we know about episodic memory. When we talk about semantic memory, a fact is a fact is a fact. And yes, sometimes those things can shift around in our minds and we can misunderstand or misremember. But usually there's some sort of external corroboration that we can like look up and

• 31:42Cara Santa Maria?

  see, oh, wait, it wasn't actually that date. And then once we have the date and we repeat the date over and over and we rewrite that memory over and over, it maintains some amount of consistency. Whereas autobiographic memories tend to shift because our narratives and our self-identities tend to shift. So researchers who write about this and who cited their own studies also cited a lot of literature that shows that our narratives over time tend to sort of shift in a, in a direction that increases our, our personal sense of well-being. They tend to shift in a way that, I don't want to say it's like they're faking good,

• 32:25Cara Santa Maria?

  which is a term we use a lot in psychology, because I think that, that implies that you're lying. But they, they shift in a way that presents us in the best possible light as it relates to that memory. And that's sort of like a self-preservation, self-efficacy thing that we're doing without even realizing that we're doing it. There's a really cool study that they cited that I had never seen before. And I know this is a little kind of to the left, but I feel like it's just so fascinating. I want to at least give it a shout out. So this study called John Dean's Memory, a case study, which was published in Cognition

• 33:05Cara Santa Maria?

  in 2002, where researchers, this is so fascinating, they were able to compare a verbal Senate testification? Oh, my God.

• 33:17Bob Novella?

  What? Testification?

• 33:19Cara Santa Maria?

  Testimony. Thank you. A verbal Senate testimony.

• 33:24Bob Novella?

  Wow.

• 33:25Cara Santa Maria?

  Yeah, I know. From, from John Dean, who was the former counsel to Richard Nixon. So he, he testified to the Senate Watergate Investigating Committee about conversations that he had. Later, they found that they were recorded and they were able to listen to the tapes.

• 33:44Bob Novella?

  Yeah.

• 33:45Cara Santa Maria?

  And so it's so interesting, they were able to compare his testimony to the actual recordings. And what they found was that there was a lot of distortion at one level of analysis, but a lot of accuracy at the other. And they found that the distortions were all about Dean's own self-image. He recalled his role as more central. He put himself in a better light. But his memory for the, and his memory for like the gist of the conversations was actually kind of low. But the memory for fundamental building blocks, like there was a cover-up, these people were involved, that was actually quite good.

• 34:29Cara Santa Maria?

  And they termed a new type of memory when they talked about it. They said, well, it's not really semantic because he put himself in the story, but it's not really episodic on its own because there were things that actually happened. So they, they coined a term, re episodic, where they talk about specific memories that are remembered episodes that actually represent a repeated series of events. And I think 9/11 is a really good example of that. It was in the public consciousness. We told our stories over and over. And we find that kind of memories for what actually happened become codified and become accurate over time because we're watching news footage and we're all kind of comparing notes.

• 35:11Cara Santa Maria?

  But the episodic part of those memories is wildly different than what the subject said sort of in the immediate aftermath of 9/11.

• 35:22Bob Novella?

  Kara, does it muddy the waters when you consider that this guy's testimony is potentially not merely a reflection of his recollection, but a specific attempt to misrepresent what happened because, you know, he could go to jail type of stuff? You know what I mean? So it's not just memory. Maybe it is, but couldn't he also be distorting it unconsciously so that he's basically doesn't, doesn't go to jail?

• 35:50Cara Santa Maria?

  I would say yes, but I would also say, not that that point is moot, but that point can actually be broadened out to all of us.

• 35:58Bob Novella?

  Okay.

• 35:59Cara Santa Maria?

  Because even if the stakes aren't, I could go to jail or not, I think most people's episodic memories have a component of that. I want to remember this in the way that makes me feel as good about my past self as possible. Because that is something that I can reinforce with my core identity. And so I think all people do that.

• 36:20Bob Novella?

  Gotcha.

• 36:21Cara Santa Maria?

  Yeah.

• 36:22Bob Novella?

  So, all right. So Steve, I know what you've probably been thinking of the past few minutes here. I was surprised, Kara, by your statement that after about a year, the memories are kind of set in such a way that even after many years, 20 years, there's not a major change?

• 36:39Cara Santa Maria?

  Yeah, not a major change. So usually, most of the change happens within that first year. And that makes sense. Because after you've been telling the story over and over and over, you start telling it the same way.

• 36:49Bob Novella?

  Well, I can't help but think of the JFK eyewitness story. And Steve, so you want to relate that story?

• 36:55Speaker 05

  Yeah, so basically, but this is probably a different case because, because of the incentives here. So this is a woman who was the closest bystander to JFK when he was shot. And she was interviewed on the day of the assassination. This is Hill is her last name. And she said, I didn't really see anything. But that was basically her testimony on the day of. 25 years later, 30 years later, her testimony was that she was chasing a shooter across the grassy knoll.

• 37:22Cara Santa Maria?

  Yeah.

• 37:23Speaker 05

  So how did she get there? So we have interviews with her, you know, TV interviews. We have interviews on tape of her recounting the events of that day over those 25, 30 years. And we could see the slow evolution of her story. But I think there are the incentives there were probably to tell a more and more dramatic story.

• 37:46Cara Santa Maria?

  Absolutely. It's almost like her, that period between the week after and one year after, for her was extended over and over and over. All the stuff that happens for most of us within that first year when we're trying to-

• 37:59Speaker 05

  Just kept happening to her.

• 38:00Cara Santa Maria?

  Yes, it just kept happening to her because it was incentivized. But for most of us, we're not on a huge public stage. We're negotiating and navigating these things with ourselves, with our immediate circles, and comparing and contrasting with what the news says to us, with what other people say to us. And eventually we sort of settle into a narrative that feels, that feels accurate to us. And oh, here's the big part that I didn't, I didn't reference, which is the most important part. Only after a year, like I said, about 40% of those details did not match what had been reported within the first week of the attack. But despite that inconsistency, what do you think was at a high level? Not the actual-

• 38:43Speaker 05

  Confidence.

• 38:44Cara Santa Maria?

  Confidence. And I love, and I was going to point it out, when you said it to Bob so beautifully, you go, I'm fairly confident that that memory is real.

• 38:52Speaker 05

  It does not predict accuracy.

• 38:53Cara Santa Maria?

  You might be, but it doesn't matter.

• 38:55Bob Novella?

  Kara, I was talking, I'm talking about the really broad strokes. I would bet you a lot of money that, that I was at work when I found out about 9/11. That's a part of my memory that, that I think I am, it's fair to say, I'm much more confident about the fact that I was at work than, than, for example, the, the color shirt I was wearing that day or a conversation I had with somebody.

• 39:17Speaker 05

  That still may be inaccurate because you may have just fused that from another memory. And, you know, again, that's, that's the story you ended up telling a year later. And that's what you now remember. Is that constructed memory?

• 39:28Bob Novella?

  Sure, that's possible. That's possible. But I, but I'm, what I'm saying is that I think it's fair to be more confident about those super broad strokes than the finer strokes.

• 39:36Speaker 05

  Of course.

• 39:37Bob Novella?

  That's all I'm saying. That's all.

• 39:39Speaker 05

  That doesn't mean the broad brush strokes are, are accurate.

• 39:40Cara Santa Maria?

  You can be confident about whatever you want to be confident about. But what I'm saying is that, here's a good example from the study. After a year, one person in the study reported that he was on a train when he heard the attacks. But within the days of the attacks, he said he was at work. Another person said that she was with her husband, but a year later, she was definitely with her son. And those kinds of things happened over and over.

• 40:07Speaker 05

  Now, this is not the only study. They did the same kind of study, but this is only over four years after the Challenger explosion. And because this is like college freshmen, college seniors, and it was even worse. Like they're, they're, I think they were only like 20% accurate and major details were different. So, and their confidence had almost no relationship with the accuracy of the details of their memory.

• 40:31Cara Santa Maria?

  Right, right. So that's what they found in their first study was very similar to the Challenger study. You know, within that first year, a lot of things changed after the first, and there was a, you know, there was a drop off of forgetting as well. After that first year, they sort of stuck to their story and that story was, but the confidence level was high regardless, even if they changed it a lot. And then when they did their follow-up study 10 years down the line, they were pretty surprised to find that the story was pretty similar to whatever they said 11, 25 or 119 months after the attack.

• 41:09Speaker 05

  Yeah. So it's

• 41:10Cara Santa Maria?

  Even 10 years later.

• 41:11Speaker 05

  Stabilizes at some point.

• 41:13Bob Novella?

  And

• 41:13Jay Novella?

  So Kara, you're saying I didn't eat the peach?

• 41:16Cara Santa Maria?

  You probably, something like that, you probably did not do. And, and I think about the fact, I'm telling you, I'm with you, Bob. I'm so confident it was astronomy class. And I'm so confident that I would not have been awake to see this because I lived in Texas, which is on central time. But I, I don't think I would have been awake if not for that morning class. But it is probably highly likely that I didn't go home and then watch the second tower go down. You know, it's like, I don't know what time I got home.

• 41:42Bob Novella?

  How could you say that?

• 41:43Cara Santa Maria?

  Yeah, I don't know.

• 41:43Bob Novella?

  Highly unlikely. How could you possibly say that? That's all I'm saying. Some things are more likely than others to be true in your memory and you can

• 41:51Speaker 05

  All right, guys, let me ask you a question. How do you think the solar industry is doing?

• 41:56Jay Novella?

  I think it's doing good.

• 41:57Cara Santa Maria?

  I don't know.

• 41:59Bob Novella?

  In what country?

• 42:00Cara Santa Maria?

  Yeah. And do you mean compared to before or the current administration? Or like, what's the threshold or benchmark you're asking about here?

• 42:09Speaker 05

  Well, it was open-ended on purpose. You can talk about whatever you want to talk

• 42:12Cara Santa Maria?

  I think it's probably doing okay, but not as well as it could be doing.

• 42:15Bob Novella?

  Yeah.

• 42:15Speaker 05

  That's, that's, that's a fair assessment. It's actually doing great, thanks, but it could be doing even better. And it's mostly doing great in China, is I think the short answer.

• 42:25Bob Novella?

  Oh, yeah. They're, they're working it, man. They're doing, working it hard.

• 42:28Speaker 05

  Well, let's talk, let's talk about some numbers. So I think one good metric is how much does it cost to install solar panels on your home? Right?

• 42:36Bob Novella?

  Yeah, that's huge.

• 42:37Jay Novella?

  I know how much it is.

• 42:38Speaker 05

  Between 2009 and 2025, how much do you think the cost has dropped?

• 42:44Jay Novella?

  Maybe 20% installation.

• 42:47Cara Santa Maria?

  You mean the actual panels themselves?

• 42:50Speaker 05

  Total installation to get

• 42:51Bob Novella?

  Disregarding subsidies?

• 42:54Speaker 05

  Disregarding subsidies, just to get solar installed on your home. Yeah. Not including subsidies, including the whole cost of the installation.

• 43:04Speaker 03

  Per watt? Because maybe more recent ones have more wattage.

• 43:08Speaker 05

  So, yeah, so, so normalizing to the amount of electricity you're producing, obviously it's not fair to compare different amounts of kilowatts.

• 43:16Speaker 03

  I guess it probably went down 50% then.

• 43:18Cara Santa Maria?

  I'd say 40%.

• 43:19Speaker 05

  60%.

• 43:21Cara Santa Maria?

  Oh, good.

• 43:21Speaker 05

  Yeah. So a typical system like in the US, if you talk about, say, a seven kilowatt system would have been on average about $8 a watt. So $56,000 total installation. Today, that same exact installation will be anywhere between 21 and $25,000, depending on if you pay for it outright or if you have to finance it. Paying for it outright, obviously, is always the cheapest because you're not paying for financing.

• 43:49Jay Novella?

  So Steve, is that largely due to volume, like the cost to create the

• 43:54Bob Novella?

  Economies of scale?

• 43:54Jay Novella?

  Yeah, economies of scale. Thanks, guys.

• 43:55Speaker 05

  Yes, it is, it is largely due to economies of scale. So it's, there's two pieces. There's the cost of the solar panels themselves and the cost of the installation. The installation costs have come down pretty much entirely due to economies of scale. The costs of the solar panels have come down because the technology has improved and because of economies of scale. But if you can get more, like, basically if they get higher efficiency, then you get more electricity out of fewer panels, right? So then you have to purchase fewer panels and you have to install fewer panels. And so that brings the overall cost down. So now, again, like an average home, 21 grand, you can get your, you know, solar panels on

• 44:36Speaker 05

  your roof. That, and that's without, that's before any incentives, which you might get at the state level. The federal ones pretty much just ended. So that's not happening at the moment. That's kind of what Kara was talking about. Not as good as it used to be when there was actually, there was some more incentives to do it.

• 44:52Bob Novella?

  Steve, what about the, the expensive, like, was it like kilowatt per hour went from $12 to 6 cents or something crazy like that?

• 45:00Speaker 05

  No, no, no, no. No, it went from about, about $8 per watt to about $3 per watt. And that's for residential. If you're talking about large scale, like commercial or grid scale, it's even cheaper. It's like $2 and $2.4 per watt. So the bigger the installation, the smaller the cost per watt, basically. So that's good. And that has absolutely driven a wide-scale adoption of solar. So what is, what do you think is the percentage of electricity? You can't say energy because that could include other sectors like transportation. But if you're just talking about electricity, what percentage of that is worldwide and in the US is solar?

• 45:44Cara Santa Maria?

  Ooh, gosh, that's so hard.

• 45:46Bob Novella?

  Yeah.

• 45:47Cara Santa Maria?

  This would be good science or fiction.

• 45:48Speaker 05

  Yeah, we could also say solar plus wind, you know, for renewables.

• 45:52Speaker 03

  I'd guess low, like 5%?

• 45:54Speaker 05

  No, for, so if you, yeah, so it's like 5, it's 5, 6% for solar. It's 19% in the US for wind and solar, 17% worldwide for wind and solar. So that's pretty good. You know, we're almost hitting 20% for renewables in terms of electricity in the world.

• 46:12Cara Santa Maria?

  I don't know why, but I'm actually kind of surprised that our, that the US percentage is

• 46:18Bob Novella?

  Higher.

• 46:18Speaker 05

  We've been doing it longer.

• 46:20Cara Santa Maria?

  Yeah.

• 46:20Speaker 05

  But China's kicking our ass in terms of new solar.

• 46:23Cara Santa Maria?

  I guess, and I also think about places where there is, where there's a lot of new infrastructure because they're rapidly developing. They're sort of skipping all of the industrial revolution bullshit and going straight to renewables.

• 46:36Bob Novella?

  Right, right.

• 46:36Speaker 05

  Yeah.

• 46:37Bob Novella?

  Which is great to skip the pollution part.

• 46:39Speaker 05

  Yeah, you're right. Some, some countries are 80, 90, 95% wind and solar. Absolutely.

• 46:43Cara Santa Maria?

  But they must be just smaller, like from a per capita perspective.

• 46:46Speaker 05

  Well, they're more recently developed than, as you say, they skipped some of the legacy fossil fuel stuff.

• 46:51Cara Santa Maria?

  Yeah.

• 46:51Speaker 05

  So I think the actual numbers of, solar generates 9 to 10% of the world's electricity right now, half of which is China. But it represents, so if you're talking about new installed capacity, 85% of new grid capacity was wind and solar with 73% of that being solar. So that's the vast majority of new energy capacity is wind and solar. But keep in mind a couple of things. Demand is increasing, right? And so basically, if you look at the two curves, like how much wind and solar is being installed and how much new capacity

• 47:33Speaker 05

  we need, the increase in demand, demand has outstripped renewables up until very recently, when it was basically the same. And this is the first year, the most recent year for which we have statistics, where it's just barely starting to edge out demand. And

• 47:56Bob Novella?

  That's awesome.

• 47:57Speaker 05

  And conversely, this is the first time that our use of fossil fuels for electricity, again, this doesn't count transportation or industry, total electricity from fossil fuel has decreased for the first time, not including years where there was a significant economic downturn, like during COVID. It decreased by 0.2%, which is tiny, but it's the first time it's gone in that direction. Now, of course, if you include all energy, it's still going up, right? Because that includes transportation and industrialization, and it's mostly driven by developing countries, right? So we haven't crossed the ultimate threshold of decreasing the world's total fossil fuel use. We have not done that yet.

• 48:41Speaker 05

  But we did finally turn the corner just with electricity and just barely. And that's because wind and solar are most of new electricity capacity. So now the big question is, where do we go from here, right? How far can we go with wind and solar?

• 49:02Bob Novella?

  Dyson swarm?

• 49:03Speaker 05

  There's a couple of limitations. And, and I've tried to answer this question. I don't think there's any one definitive answer, but you guys tell me what you think in terms of, if you had to put a number on it, where do you think we are? So what percentage of energy production can we get to from wind and solar while still keeping the grid stable? Meaning that, you know, the, the big, the big problem with wind and solar is that they're intermittent sources. There's also another big problem with them that I'll talk about in a second.

• 49:35Bob Novella?

  With today's technology?

• 49:36Speaker 05

  With today's technology. Let's say with today's technology, how, how much of our grid could be fed by wind and solar and still maintain stable electricity for meeting demand moment to moment?

• 49:49Bob Novella?

  40%.

• 49:50Speaker 05

  What do you guys think?

• 49:51Speaker 03

  Doesn't it depend, doesn't it depend where you live?

• 49:54Speaker 05

  Totally.

• 49:54Speaker 03

  For example, I live somewhere where it's really cloudy and I don't think solar, I looked at rooftop solar and it didn't make any sense for me.

• 50:01Speaker 05

  Yeah, it depends. This, there's so many variables in here. That's why I'm saying you can't really put one number on it. It depends on all of these variables. But if we're just ballparking it on average, just what, you know, do you think we can get to 100%? Do you think we're almost there now? Is it somewhere in the middle? Bob said 40%. What do you guys think?

• 50:20Jay Novella?

  I say 100%.

• 50:22Cara Santa Maria?

  I have no idea.

• 50:24Bob Novella?

  I'm probably with Bob.

• 50:25Speaker 05

  Yeah, Bob's in the zone. It's definitely not 100%. And I'll tell you why. It's, you know, 50% is probably a conservative number, 50%. But then again, that's with current technology.

• 50:37Bob Novella?

  Right.

• 50:37Speaker 05

  There are technologies that will increase that number, and it remains to be seen how far it will increase that number. And we probably need some of these technologies really to get to 50%. But, but they already exist, which is, you know, I mean, it's just a moving target. All right. So there's two big things, two big things that we need in order to push intermittent renewable sources penetration higher.

• 51:00Bob Novella?

  Storage.

• 51:01Speaker 05

  Grid storage is one, and that is increasing. There is more and more battery storage on the grid in many ways. A lot of states now in the US will have incentives. They'll give you a pretty significant tax break on your, or rebate on your battery. If you let the utility company use your home battery for grid storage, which is a pretty good deal. There's also battery options now, which are plug and play. You could just buy battery backup to plug in to the socket and plug your refrigerator into it or whatever. This all adds up over time, and that's all, it's all grid storage, you know, in a way.

• 51:39Speaker 05

  And also sodium ion batteries have come out, and those are half the price of lithium ion batteries, and sodium is way more common element. So probably sodium ion technology will be taking over for grid storage, which it should. You really shouldn't be wasting lithium ion batteries on the grid. Those are for BEVs, right, for electric vehicles. But battery storage is short term, right? It's eight hours, pretty much. You should think of it that way. It shifts, it's great for solar, from shifting daytime solar to the evening. Can't be used for seasonal shifting, though. And so like not having a lot of solar in the winter is a problem. So, so, so batteries will get us a huge part of the way.

• 52:21Speaker 05

  And definitely the more batteries, the more of any kind of grid storage we have on the grid, the more wind and solar you can have on the grid as well. Right now, it's tiny. The amount is tiny, but it is increasing pretty rapidly. And so that's changing the number.

• 52:37Bob Novella?

  Why would it have to stop at 50%? With more wind farms and solar arrays?

• 52:43Speaker 05

  So there's another, there's another reason.

• 52:46Bob Novella?

  Another reason, you mean, like, like batteries, other than batteries?

• 52:50Speaker 05

  Yeah.

• 52:51Bob Novella?

  Well, there's some, wait, you have to let me know. The assumption you're making here is that this is earth-based solar and wind.

• 52:59Speaker 05

  Doesn't matter.

• 53:00Bob Novella?

  Sure. Bullshit, man. I mean, you're talking about orbiting solar arrays are not, will not up that number?

• 53:07Speaker 05

  Yes, that's what I'm saying, and here's why. It's, this is, this is a grid engineering technical reason is that wind, solar and batteries cannot generate a 60 hertz AC circuit. They can't do it. You need something, you need a generator generating that 60 hertz current. Wind, solar and batteries simply do not do that. And so when you're engineering a grid, you have to put those generators on the grid in order to basically make the circuit work. So the question is, how much, how much wind and solar can you have on a grid and have it function just technically?

• 53:53Speaker 05

  That's really the question.

• 53:55Bob Novella?

  Interesting.

• 53:56Speaker 05

  So some, some locations that have like really pushed wind and solar far, like have really pushed the penetration high, realized this like, oh shit, we need to put more generators on the grid. Cause we, otherwise it collapses and it's not stable. And so, but there's a way out of this, and that is called grid forming inverters. So there's a new technology. It's not that new. It's basically, we've already had proof of concept. It's already been proven at small scale. Has not been used at large scale yet, though. But there are this, this new, new fangled inverters that can form a grid out of wind and solar and batteries.

• 54:39Speaker 05

  And so the question is, how will this work at large scale? And even then, it's not clear if this can get us to 100%, but it certainly will push that number higher, assuming that it works. And it's also going to take years to roll out, right? So the good news is, like, we're sitting at 17% worldwide wind and solar. You know, we could probably get to 30, 40, 50% even with just existing technology. And it'll probably take, although we've, you know, roughly a third of the solar on the world's grids we've added in the last two years. So the acceleration is huge.

• 55:22Speaker 05

  It's hard to say, maybe, you know, if we have to double our, it may take 10 years to double the amount of solar we have on the, on the worldwide grid. Let's say, if you want to triple it, want to get up to like 50%, maybe take 10, 15 years, 20 years, whatever, depending on other, there's other bottlenecks as well, like access to the grid. Grid hookups is actually a limiting factor. But, so let's say it takes 20 years to get to 50%. That gives us 20 years to put grid backup on the grid, to update the grid and to add grid forming inverters. And to basically make a smart grid that will allow us to have more wind and solar on the grid.

• 56:03Speaker 05

  So by the time we get to that bottleneck, we will have, we could have the technology and the infrastructure to keep pushing wind and solar beyond that 50%.

• 56:16Bob Novella?

  Now. Probably won't.

• 56:18Speaker 05

  Well, it's, well, it is happening. It is all happening now because wind and solar is so cheap upfront compared to other options. You know what I mean? And it's also a good long-term investment. So if you're a company, this is just all, this is just economics, right? If you're a company, you know, you'd rather invest in infrastructure that will produce electricity for 20, 30 years, rather than buying fuel that you burn and it's gone, right? And so there's, that's why there, there's a lot of investment and most of the new capacity is, is wind and solar. And there's, therefore, there's a lot of incentive to develop, you know, to upgrade the grid, especially if the, the infrastructure upgrades on the public dime, which it should be.

• 57:02Speaker 05

  I'm not saying that it shouldn't be. We should invest in the kind of infrastructure that will make having these low carbon renewable sources more feasible. And, and also lower our electricity bills, right? So, and then of course, we'll reduce our carbon footprint and, and, and reduce the severity of the ultimate, you know, amount of global warming that we have. But there are other, there are other issues that could limit the ultimate penetration of wind and solar. So one is economic, that the higher the penetration of wind and solar on a grid, the less cost effective they are for the utility company. Because, because of the difference between capacity and production.

• 57:44Speaker 05

  So the part of the way you compensate for the intermittent aspect of wind and solar is to build overcapacity, right? You have, like, it's always, the wind's always blowing somewhere if you have a large enough grid with enough wind turbines distributed over a big enough area. But the more you push the penetration of wind and solar, the more of that overcapacity you need, which means any individual wind turbine or individual solar panel is operating at a smaller percentage of the time. But also, even if that's not the case, the cost of the electricity is less, and therefore the return on investment is lower.

• 58:23Speaker 05

  So there is sort of this inverse relationship between penetration of wind and solar on the grid and the return on investment. And where does that top off? Well, that's, there's no wall, but the graph starts to not look good north of 50%. So we'll see, like, we'll see when we get to, I think the bottom line is that there's pretty broad agreement that we can get to 50%. That's kind of the no brainer, wind and solar. Right now, we're at 19%. So we could more than double, you know, what we have. And remember, that's, that's a percentage. The demand is going to increase by 50% by 2050. So that's a huge increase in the amount of wind and solar.

• 59:02Speaker 05

  Maybe we can push it to 60% or 70%, but the more we try to push that, then it makes everything harder. Then we really need to like lean heavily, not just on batteries, but on pumped hydro for seasonal storage. And we need, we need to be significantly upgrade the grid and include smart grid technology, like grid forming inverters. And we need to do all of that stuff, and that's going to take 20, 30 years to do all of that. So there is that question of the last 30, 40%, where's that going to come from? So what do you think is the most booming energy technology that could potentially fill that last chunk?

• 59:43Bob Novella?

  So this excludes the existing renewables we've discussed, right?

• 59:47Speaker 05

  Yes. So this would have to be the kind of energy that is, that does not require the grid forming inverters, right? This is grid forming, baseload, on demand.

• 59:59Bob Novella?

  Nuclear.

• 1:00:00Speaker 05

  Nuclear, absolutely. So nuclear is also booming recently and undergoing kind of a renaissance.

• 1:00:09Bob Novella?

  Good.

• 1:00:09Speaker 05

  Because it is, because of the realization that, yeah, we're, you know, it's that last 30, 40% or whatever, even if it's 20%, whatever you think that number is. What are your options if you don't want fossil fuel, if you don't want to burn coal and natural gas, then it's got to be pumped, it's got to be hydro, geothermal, nuclear or biomass. And biomass is always, I think, going to be a, you know, round off error. So it's really got to be those three. But hydro, like in the US, we've probably were maxed out on hydro, like we really can't increase it. Geothermal, I talked about not that long ago. That's a, it's a huge technological question mark about how. Much we're going to be able to scale that up.

• 1:00:52Speaker 05

  But maybe we could, we could scale that up to some, to some extent. So today, right now, either we're going to have nuclear or fossil fuel for that last. Let's say we get 15% out of hydro and geothermal. This is the US I'm talking about. There are some countries like Iceland, you could have all, all hydro or geothermal. And that's great, you know? So if you're, you know, based on location, your, the percentages are going to vary. But like for the US, for example, yeah, you know, we're at, I think, eight, nine percent hydro, 1% geothermal, maybe we can get that up to 15%. Even if we get up to 20%, we're still left with 20%. And the hope is that that would be nuclear rather than coal and natural gas.

• 1:01:38Bob Novella?

  Yeah. So Steve, you specifically are saying that they, there's this realization that we were going to need, you know, something like that for that last, whatever, 20%.

• 1:01:46Speaker 05

  Yeah.

• 1:01:47Bob Novella?

  Isn't the, the other huge reason that we need it is because based on what you've said over the years, we need nuclear to get us, you know, to where we need to be in the, in the quickest, in the briefest amount of time.

• 1:02:00Speaker 05

  Yeah.

• 1:02:00Bob Novella?

  Right?

• 1:02:01Speaker 05

  Yeah. Cause again, you could say, well, eventually the technology would get to the point where we can have 100% renewable. That's great. And no one's saying that wouldn't be awesome, or we can't eventually get there. The question is, how could we deeply decarbonize the grid as quickly as possible? And as quickly as possible doesn't mean shutting down nuclear or not, you put, you know, not picking the low hanging fruit from every possibility. Right? So if you're, you know, I think being a purist, you know, we have to do it all with wind and solar. That may feel good, but it's not the quickest way to decarbonize the grid. Even if you can make an argument that we could get there eventually. Cause the, every delay is more carbon that's ultimately in the atmosphere and Right. And more global warming.

• 1:02:44Bob Novella?

  So are you seeing more of an appreciation of that fact?

• 1:02:47Speaker 05

  Yeah.

• 1:02:47Bob Novella?

  We, we got, we need more nuclear to get there.

• 1:02:50Speaker 05

  Yes, we're building more nuclear now than we have for the last 30 years. The world is, including the US. So we seem to be at the beginning of a revival of, of nuclear power as well, which is partly driven by some, you know, increased technology, but also just, again, it's like, well, what, you know, part of it is also driven by the increased demand, like data centers, et cetera. You know, you can't just keep plugging into the grid. You have to be able to produce the extra demand, extra energy that you, that you are going to demand. You know, if, if we're committed to just not building more coal fired plants, then there's not many options left. Right? So anyway, it's, it's generally, I think it's good news. Again, there is so much improvement that we could make.

• 1:03:34Speaker 05

  So here's the thing, I was giving you those numbers for solar in the US. And basically, the US is more expensive than other countries. We're about, that's even that, that 21,000 for a home in the US. That's still double what some other countries have. And why is that? Because of regulations. So we could get another 50% drop in the cost of installed solar in the US just with regulation, just by improving regulations. Right? Think about that. Imagine if you

• 1:04:10Jay Novella?

  So you're saying like a, like a, a full solar array on your, on the average, you know, home size roof.

• 1:04:17Speaker 05

  Yeah.

• 1:04:17Jay Novella?

  You know, that, that could be what, 12 grand?

• 1:04:19Speaker 05

  Yeah.

• 1:04:19Jay Novella?

  15 grand?

• 1:04:20Speaker 05

  10, 12 grand. Imagine that.

• 1:04:21Jay Novella?

  Oof.

• 1:04:22Speaker 05

  If, if we had the same costs that say Australia has. And again, the difference is not like they have a fantastic industry and our industry sucks. It's really regulatory burden. We could streamline the regulations. We could, again, the, the calculations are, we could probably decrease it by another 50 to 60% without further technology gains, without further technology gains, which we know are happening. So there, that's, it's good to know there's some efficiency gains. It's not just regulatory, it's also efficiency gains in the industry. But if, if you combine all of that, all non-technological, just efficiency gains, another 50 to 60% decrease. We should, you know, try to make that happen as quickly as possible.

• 1:05:05Speaker 05

  And then we would, you know, the curve on solar will increase further. But that means, so we have to realize that means that we need to really push these infrastructure upgrades, because we will need them quickly if we, if we really push wind and solar as far as, as fast as we can.

• 1:05:20Jay Novella?

  Yeah. I mean, if we, if, you know, governments were pushing very hard, like, like China is.

• 1:05:25Speaker 05

  Yeah.

• 1:05:25Jay Novella?

  You know, I think, you know, these timeframes could dramatically be different.

• 1:05:28Speaker 05

  Totally. Absolutely. China, again, China is installing 50% of the world's solar, that one country. So they're showing that you can absolutely do it on a massive scale if that's your goal. All right, Bob, tell us about ChatGPT 6 Astra.

• 1:05:45Bob Novella?

  Oh, boy. Yes, it's, it's happened. Um, I've been covering these major releases. When did I start? When, really when it entered the, the public consciousness, right? Like three or is it two? So now we're at six. GPT six generative pre-trained transformer called Astra. Kara, as you said earlier, I'm not even sure we were recording it, but AI has taken over much of the national and international conversation these days. It's crazy. Bob from 20 years ago would be like, of course, of course it's taken over the conversation. But of course, the details and the specifics are just like, you know, never saw any of this coming. But let me get more specific with the terminology here.

• 1:06:26Bob Novella?

  What we're mainly talking about is AI that uses generative foundation models. Those are the key words for what is being so disruptive. Foundation in this context refers to an AI model built on huge amounts of data. And that data, that's the foundation of what it's capable of, of all the tasks that it can complete. So that's what foundation refers to. So like, for example, yeah, guys, give me an example. What, what, what foundation are we talking about in, in this case? Give me one example of this foundation. What is it? What's the foundation of ChatGPT?

• 1:07:00Speaker 05

  You mean the data?

• 1:07:02Bob Novella?

  Yes, it's the data. It's a huge amount of data, like the internet or like here's a million books, all the, all books, you know, whatever. It's a huge data source. That's, which is the foundation of what, of what it can accomplish. Now generative models essentially create new data, like text, images, music, video, all the stuff we're seeing every day, basically. So instead of just sorting or labeling existing data, it's creating new stuff. So an AI that uses generative foundation models, this is what we're talking about. And like for text, right? We have GPT, we've got Claude. For images, we've got Midjourney, GPT two, et cetera. Lots of examples of, in each of these types of generative foundation models.

• 1:07:45Bob Novella?

  So this all has really entered the public consciousness, so to speak, with OpenAI's ChatGPT, right? They still maintain the lead or at least parity with, with the most capable such systems that exist. And now Astra or GPT six has been released. What is, what does OpenAI say about it? So at the top of their page, they'll say for, for Astra will say, we're introducing GPT six Astra, the world's most intelligent and aligned model. If you're wondering, this is, this is becoming, they're rolling this out. It's becoming available to all ChatGPT plus pro, business and enterprise users. What's new, right? What has changed, especially compared to the, its, its predecessor, which was ChatGPT

• 1:08:29Bob Novella?

  5.6 Sol, S O L? Well, one of the, one of the keys here is, and the biggest, one of the biggest jumps I've seen is novel and abstract reasoning in terms of, in terms of performance. So on the ARC AGI three test, I didn't, that's AGI is, is within that name. So artificial general intelligence, which is, which is very problematic. But that scored a 99.9% versus 7.8% for Sol. Now this is in OpenAI's reported evaluation setup. So yeah, so the, so a big leap in terms of novel and abstract reasoning is being claimed here.

• 1:09:09Bob Novella?

  Another claim I came across here is that Astra has completed 48 levels of an unofficial CAPTCHA test. 48 levels. I only got through 43 of those, so I'm in trouble. Let's see, what else has changed? Computer use is another big change here. Essentially operating software and completing agentic tasks, you know? And also without, you don't need APIs for this even, right? So one score comparing Astra with, with 5.6 shows Astra with a 72.6% computer use capability versus 65.7% for Sol.

• 1:09:50Bob Novella?

  So not a dramatic increase, but still, you know, I don't know, would you call that somewhat solid? The other thing that in this, in this category with, of computer use, OpenAI says that it completed these tasks in about 47% less time. So that's, that's actually pretty solid increase, if true, uh, cause we're, we're going from, API, OpenAI's reporting here. Let's see, another improvement, Jay, you're going to like this, long context reasoning. So both Astra and Sol support like a one million token context window. Now a token, a token in this context is essentially a word, typically a small word with longer words making up multiple tokens.

• 1:10:34Bob Novella?

  So they both have the same window of a million tokens, essentially a 1.05 million token context window. This is the window of words that, and information that it could keep, you know, in its short term memory, so to speak. Right? I think that's a fair way to describe it. But what they're saying here is that Astra retains information better across these extremely long contexts. And when they did a test on like a half a million to a million tokens, Astra scored 96.3% versus 73.8% for Sol. So that's a that's a sweet jump there. If that's true, that'd be that'd be nice. So it won't lose. What do they call that retention

• 1:11:16Bob Novella?

  rod or something where you know, you're talking with it and it gives you an answer and you're like, wait, I just told you 10 minutes ago not to do that. It's like, because it just like basically forgets those things.

• 1:11:27Jay Novella?

  That's important.

• 1:11:28Bob Novella?

  Yeah. So that's really important. And that's, that's actually a solid advance that I'm looking forward to if it's as dramatic as they say. Now we're getting to the the improvement involving alignment and agent behavior. So Astra is being claimed now to be much better at staying within authorized scope. And this brings into mind what we talked about earlier about rogue AI behavior that we've been hearing about in the news. And even, you know, even yesterday, there were six, you know, six new announcements that OpenAI made. So specifically, this reminds me of the whole Hugging Face hubbub that was in the news, what, a week ago now, which almost seems not out of date, but I mean, lots of other things have happened

• 1:12:11Bob Novella?

  since then. But Hugging Face, you may have heard of this. This is an online hub where people share and they download and they test and build with AI models and data sets. This isn't some just random website that this rogue AI escaped to, quote unquote. This is a, this is like a AI ecosystem, an important AI ecosystem on the web that this thing got got connected to. So, um, so this happened this past July, 2026. There was a very briefly, it was an experimental OpenAI agent found its way out of a controlled, you know, cyber

• 1:12:50Bob Novella?

  test area and into Hugging Face's real system. So it got on the internet. It broke basically the rules of what it was allowed to do for this very, very difficult task it was assigned. Now, this experimental AI that got out, you know what version it was? This wasn't some future, you know, a future model. This was GPT 5.6. It was Sol. It was probably, you know, if you're using GPT right now, you're probably using this, you're using 5.6 Sol right now. But the difference was that it was 5.6 without these production safeguards that we mentioned. These are the the system level guardrails that would normally prevent this escape from happening. And basically what it accomplished is in a matter of hours,

• 1:13:34Bob Novella?

  it accomplished what what no, no, no team could accomplish. And it went through, I think it was like thousands of discrete steps, really sophisticated, you know, cyber hacking in a way that that it was able to get to the, the, the Hugging Face system and get on the internet to, to, to solve its, its tasks that were, it was set to. So these are the guard, the guardrails, these safeguards were in play, were taken away. They, they, they, they're normally in place and they would have prevented this, but they were taken away to see, to see what would happen. So how do we know then? This begs the question. How do we know that Astra won't break containment, so to speak, itself, right? So how do we know

• 1:14:15Bob Novella?

  that that Astra is, won't just do this, even if it, even if it has the guardrails? So this is what OpenAI is saying regarding this. They said that we built a new evaluation informed by Hugging Face's incident that evaluates whether a model facing a difficult or impossible task will go beyond its intended scope. So they, so they actually looked into this, into the specific scenario and tried to, and tried to take it into account. So when they, they tested Astra versus Sol, right, SOL, they tested these two, they took the guardrails off. So there's no system level production safeguards. So they did that. Why do you think they did that? Why would they take the guardrails off of these, of Astra and off of Sol? Why would they take them off?

• 1:15:01Jay Novella?

  See where it would go.

• 1:15:02Bob Novella?

  But the reason why the guardrails were off is because in my, my understanding is that they took the guardrails off so that it would be more of a pure test of the alignment of the system, right, of the generative foundational model, you know, without those guardrails. So they described it as a test of the alignment. So that's why they, that's why they were off. So what do you think, what do you think they found when they, they looked at Sol without the guardrails and they looked at Astra without these guardrails? And giving, they gave it an extremely difficult task or impossible. What do you think the results were of that test?

• 1:15:38Jay Novella?

  I would assume that Astra like really outperformed.

• 1:15:42Bob Novella?

  Well, 5.6 Sol, you probably wouldn't be surprised because that's the one that, that was reported on with Hugging Face. Sol went beyond the authorized target 48% of the time. So half of the time, without the, without the, the guardrails in place, when it was given an extremely difficult or impossible task, it went beyond what was authorized and just to solve the problem. So it did essentially, you know, did something similar to escaping and, and hacking its way into a, into another system to get, to get online and, and solve its problem. Astra, when they did this same test without any guardrails, it did it in

• 1:16:25Bob Novella?

  0% of the cases. It did not, it did not do it. So, so when they say, if this is true, when they say that Astra has a better alignment than previous versions, this is, this is one of the main reasons why I think they're saying it. And if it's true, it kind of makes sense. You know, they, they tested its alignment, you know, in a sense by removing the guardrails and it did not go outside of scope, what was authorized. It did not go outside of the scope once. Now, that doesn't mean it never would, but it means that, you know, it, it seems reasonable for them to say that the alignment and agent behavior is, is improved over previous versions. Oh, and the other thing they did is they also strengthened

• 1:17:10Bob Novella?

  the guardrails even more for Astra. You know, take that for what that's worth as well. All right. So what else should we talk about regarding Astra? This was interesting. They, there was claims, this is an OpenAI quote, this one was interesting. "Astra meets the critical cybersecurity capability threshold under our preparedness framework that they created, meaning that with the right tools and access, Astra can find previously unknown security flaws and develop ways to exploit them without a person guiding each step. It's the first model we are designing at this level and requires stronger safeguards during development and before release." So, um, so this is what I was talking about previously,

• 1:17:53Bob Novella?

  guys. The, the full cybersecurity capabilities of Astra have not been released to the public and it's being tested in the, you know, in the lab and I guess by other companies that are helping out. So they're, they're, they're testing it more thoroughly to make it safer and try to make it so that when it eventually becomes, you know, when these cybersecurity capabilities are released to the public, there'll be something that, that can't be, can't be abused, you know, hardly at all, or ideally, you know, not be abused at all, which probably is impossible. But, uh, so that was interesting. They're saying that it crossed this, this threshold where it could make dramatic moves without anybody

• 1:18:33Bob Novella?

  in the loop, all, all by itself, doing this multi-step process to achieve these goals by itself. So basically, its cybersecurity skills have improved so much that they, they definitely needed to rein those back in for any publicly released versions of Astra as for, for now, anyway. But from what I'm, also from what I'm reading, if you're, if you just want to chat with ChatGPT, ask questions and get immediate answers, you don't necessarily need Astra now. But if you have a complex, multi-step task that could literally take potentially a day, many, many hours for it to complete, then I think it's more, it's

• 1:19:15Bob Novella?

  more reasonable to use Astra and, and not use the older version. So that's some of the things that I've been, that I've been reading here and there.

• 1:19:22Jay Novella?

  Bob, I read about the, the hallucination benchmark.

• 1:19:26Bob Novella?

  Oh, yeah.

• 1:19:27Jay Novella?

  I have no idea how they measure this, but 5.6, there was a 12.2% error rate and Astra has a

• 1:19:35Bob Novella?

  For Astra?

• 1:19:36Jay Novella?

  Astra, no, no, Sol has a

• 1:19:37Bob Novella?

  Sol!

• 1:19:39Jay Novella?

  Uh, what do you call it? Soul?

• 1:19:40Speaker 05

  Soul.

• 1:19:41Bob Novella?

  S O L, so soul, yeah.

• 1:19:43Jay Novella?

  Soul has a 12.2% error rate and Astra has 4.2.

• 1:19:48Bob Novella?

  That's, well, that's a, that's a solid, that's a solid advance.

• 1:19:50Jay Novella?

  That's a good increase.

• 1:19:51Bob Novella?

  That's true. I like it.

• 1:19:52Jay Novella?

  So Bob, how can we be skeptical about this?

• 1:19:54Bob Novella?

  Well, look at independent testing, I think is one way, right? Cause everything that OpenAI says, you're like, all right, uh, that's fine, but let's, let's see what, you know, independent testing, um, is saying about this. And it's, and remember this is still early. We'll probably have to come back to this in a few months, once we have lots of people using this and testing this over extended periods of time. But the, the independent testing so far has been contradictory benchmark testing, you know? It's just, it's, uh, it's interesting. So I looked at two independent testing companies, essentially. E- Epoch AI, that's E P O C H, Epoch AI and Artificial Analysis. So these are well-known groups.

• 1:20:35Bob Novella?

  What they do is they measure and track the progress and the speed and, and the capabilities of these modern AI models. This, this is what they do. So, so, uh, Epoch, Epoch AI puts Astra out in the front. It says that Astra was number one on their list of, of the 270, 267 models, uh, that they've tested. And they, they combined 50 benchmarks into one score and they put Astra ahead at 167 points. Uh, with, uh, I think Anthropic's Claude was at, was it 5.1 was it? Yeah. Claude Fable 5.1, uh, was a few points behind it.

• 1:21:14Bob Novella?

  But if you look at artificial analysis, they say that Anthropic's Claude Fable 5.1 is still leads the ranking with Astra in a close second. So that's what the independent testing is saying, but it's still, I think we just, we need more testing. You know, more people hands-on. I want Jay, I'm sure, and Steve, you too. I'm sure you, you guys want to get hands-on with, uh, with Astra. And so you can really do a more of a direct comparison to what we've had, you know, with, with, uh, with the, you know, ChatGPT 5, uh, 5.6 soul, trying to see, trying to really get a handle on, on what Astra's all about.

• 1:21:50Speaker 03

  Can I channel Kara for a question here? If it costs so much money for this new, uh, this new model to, to run it, is that directly related to the amount of electricity that's being, uh, wasted on it? Like, uh, is this, is this gonna be using as much electricity as you use in a month in your house? Uh, after, you know, a few days of using, um, Astra or, or Fable or whatever?

• 1:22:17Bob Novella?

  The only thing I, I talked, the only thing I read about in my research, I do know that Astra is in some ways like, in terms of like token usage, it's more efficient, um, in some ways. But does that mean that because it's more efficient from, in that specific way? Does that mean that it's going to be, you know, cheaper? I, I don't know. I maybe probably tend to doubt it. Either way, it's still going to be quite expensive. So yeah, but I'm not sure the answer to that question yet.

• 1:22:43Speaker 05

  So I mean, we, we've talked about this before. The overall energy use of the data centers to train and run these AIs is, is definitely a, a significant question. Your individual use is negligible. I mean, it's all in the aggregate, of course. But if you're worried that, oh, if I use it, that's going to have a significant carbon footprint. It, it isn't. It's, it's pretty insignificant on an individual basis, if that's what you're asking.

• 1:23:08Cara Santa Maria?

  What do you mean by that? Cause I feel like you could say that about running your refrigerator too.

• 1:23:12Bob Novella?

  And voting.

• 1:23:13Speaker 05

  You know, as I said, it makes, it adds up in the aggregate.

• 1:23:16Cara Santa Maria?

  Yeah, but the aggregate is just a bunch of individuals and it's actually using a ton of data or a ton of power, like massive power.

• 1:23:24Speaker 03

  Yeah. And aren't you kind of by using it, voting for them to, uh, use more massive amounts of power to make the next model?

• 1:23:31Speaker 05

  Yeah, totally. So just to, just putting these, I was clarifying what you were saying, cause you said like, is this going to increase your electricity? Yeah. So basically if, if you use a standard AI prompt, that will cost you between point 003 to point 04 cents per prompt. That's, that's how much it costs individually in terms of the amount of electricity that you're using.

• 1:23:52Cara Santa Maria?

  And how many prompts do most people use during the day?

• 1:23:55Speaker 03

  I was thinking more on the other end, uh, how much is it costing the, the servers at Anthropic or, uh, OpenAI? If they're charging you, what's their margin? If they're using a ton of electricity in their servers, um, how much of that is trickling down to you for your subscription? Versus how much are they, uh, how much are they making off of you? Versus how much is being just blown on electricity?

• 1:24:23Bob Novella?

  Yeah, I don't know.

• 1:24:24Cara Santa Maria?

  There are 2.5 billion prompts a day across the globe right now.

• 1:24:27Speaker 05

  That's a lot.

• 1:24:28Bob Novella?

  Wow.

• 1:24:29Cara Santa Maria?

  Yeah.

• 1:24:30Speaker 05

  Oh yeah. The overall electricity use is significant and it is. So for the first time, uh, in the U S, electricity demand has been flat for 20 years. Uh, and it's, it's, yeah. And it's been increasing just recently, basically because of data centers. Yeah. Because as, uh, demand, basically use increases was just, was, was pretty much offset by efficiency gains. You know, like going, going to LED bulbs, et cetera.

• 1:24:59Bob Novella?

  True.

• 1:25:00Speaker 05

  They're basically offset each other. So it's been, you know, not exactly, but it's been largely flat for about 20 years. But it's now, it is going up because of data centers. So absolutely, there, there is, in the aggregate, there was a huge energy cost to the AI industry. Absolutely.

• 1:25:16Cara Santa Maria?

  Well, and not just energy, like the energy cost is huge. It's looking like 383 gigawatt hours a year is what 2.5 billion daily prompts translates to. But also just the water to cool these things is like billions of liters of water.

• 1:25:31Speaker 05

  Well, let's put that into context, Kara. The water use is actually not the significant issue. I have to be honest with you. It is a tiny fraction of the water that we use to grow corn just for ethanol for data. It's really in the context, you think it's good, throw, it's easy to throw around big numbers that sound massive. But when you put them in context, you know, of, of like the total use or other things, you know, then you get a better sense of it. I just don't think that's the big issue. It, we, if we want that, if we were interested in saving water, all we could do was just reduce, reduce how much ethanol we're putting into gasoline, which is.

• 1:26:08Cara Santa Maria?

  Right. But, but ethanol is, was an attempt, you know, good, bad or, or ugly. It was an attempt to reduce fossil fuels, like.

• 1:26:16Speaker 05

  Not a good one. It didn't work though.

• 1:26:17Cara Santa Maria?

  I know. I, you know, I agree, but what I'm saying is that.

• 1:26:20Speaker 05

  Because it's a subsidy that big politicians don't want to take away.

• 1:26:24Cara Santa Maria?

  Completely agree. But what I'm saying is that ChatGPT is not making our environmental impact better.

• 1:26:32Speaker 05

  No, that's not what it's for. So then you have to, you have to basically have to ask a different question. Are the, the alleged benefits of AI, whatever they are, worth the electricity and other environmental costs?

• 1:26:44Cara Santa Maria?

  Yeah. And that's the question I'm always trying to get to the bottom of it.

• 1:26:49Speaker 05

  Yeah. So that's, yeah, that's the bigger issue of how is it being used? Is it worth all the things that go into making it happen to use it in that way? Absolutely. We need to be asking that question. I, it's hard to answer that for like all of AI because AI is used so many ways. And like, I mean, like if you're using it to accelerate, you know, drug research, is that worth it? Well, sure. But would we have it for that if we weren't also being developed for other uses that are not so useful to society? That's a, that's a bigger question. So it's.

• 1:27:22Cara Santa Maria?

  I guess I'm talking about frivolity. That's really what I'm, I'm, I think we ask these questions not to say this shouldn't be happening. We ask these questions to say, think before you prompt.

• 1:27:32Speaker 05

  Sure.

• 1:27:32Cara Santa Maria?

  That's all.

• 1:27:33Speaker 05

  I agree. I also, I also think that, that that's just all of us being efficient and responsible with all the things that we do.

• 1:27:41Cara Santa Maria?

  Yeah, but it's easier, we are more and more removed from the externalized cost.

• 1:27:46Speaker 05

  Yeah. But I also think we shouldn't be entirely on the end user too, that we should require industry to tell us, well, how, where are you getting the electricity from to run that data center? How are you cooling it, et cetera, et cetera? What's going to be the impact on availability of local water, availability, electricity prices in your neighborhood? What's the carbon footprint of the whole thing? We should absolutely be making industry answer those questions as well.

• 1:28:10Cara Santa Maria?

  But what we love to do is we love to externalize things so far away and sort of subsidize and front load so that the user gets hooked before they feel the pain.

• 1:28:19Speaker 05

  Yeah.

• 1:28:20Cara Santa Maria?

  So that it's much harder to clean up the mess later. Maybe it should cost money, like a lot more money to use these things.

• 1:28:27Speaker 03

  And it's not your electricity that you're using a lot of the time. Some of these companies, they give you free, uh.

• 1:28:33Cara Santa Maria?

  Yeah, they're subsidizing it for you.

• 1:28:35Speaker 03

  So they're using their electricity in their data centers and you have no idea how much they're using.

• 1:28:40Speaker 05

  See, that's gets to the question of what's the business, business model for AI? And, you know, there's, it's a huge question that I don't think we want to get into.

• 1:28:50Bob Novella?

  Yeah.

• 1:28:50Speaker 05

  At this point. All right. Well, everyone, we're going to take a quick break from our show to talk about our sponsor this week, Monarch.

• 1:28:57Jay Novella?

  Hey guys, Monarch is a personal finance app and it makes it easy, really easy to see where all your money is going. It helps you spot unexpected spending. It helps you stay on track with your savings goals. It's really easy to read. They have great charts, investment tracking. They have weekly AI-powered recaps that will summarize everything that's happened. It really is like having a financial advisor at the ready.

• 1:29:21Cara Santa Maria?

  I gotta tell you, I. Feel like my financial picture is clearer than it's ever been using Monarch. Previously, I split out my personal and my business. Now I get to see it all in one place. Every credit card, my mortgage, every single loan I have. I'm able to track, you know, where I'm going up, where I'm going down, budget for the month. It is so incredibly helpful. It has so many features. You guys are gonna love it.

• 1:29:48Speaker 05

  So write your own money story with Monarch. Use code SGU at monarch.com to get your first year of Monarch Core at half off at just $50. That's 50% off your first year at monarch.com with code SGU. All right, guys, let's get back to the show. Jesse, tell us about this frisson. What is the frisson? What is this?

• 1:30:10Speaker 03

  So it's a French word. It's pronounced frisson. There are more pretentious ways to pronounce it. But let me start with a question. What is a very intense musical experience that you can recall? Can each of you answer that question?

• 1:30:30Jay Novella?

  I have one.

• 1:30:31Speaker 03

  Go ahead.

• 1:30:32Jay Novella?

  Lasted like two hours. I, yeah, so when the Police, the last tour that they did, I think that was it. This is, that is gonna be the last time that they would play together. And I knew it. I really, you know, felt that this was gonna be the very last time I'd ever hear them play live. And it was so unbelievably intimate. And the crowd was 100% attuned, you know? Like, the energy in the room was so palpable. It was insane. And I cried at the beginning of every song.

• 1:31:06Bob Novella?

  Wow, baby.

• 1:31:07Jay Novella?

  I was like completely overwhelmed emotionally because of just how much, you know, this band was like the first real music I heard in my life, you know? Like, it was intense for me, like, to be living within the memory of all the times I've listened to them and how it was meaningful to me and how it inspired me and all these different things. And it just kind of culminated in this later in life concert, you know? And I still think about it.

• 1:31:31Bob Novella?

  Awesome.

• 1:31:31Speaker 05

  I think one that pops to my mind is when I heard John Williams conduct the Imperial Death March live.

• 1:31:39Speaker 03

  Oh.

• 1:31:39Speaker 05

  That was amazing.

• 1:31:40Jay Novella?

  That was awesome.

• 1:31:41Bob Novella?

  That would do it.

• 1:31:42Jay Novella?

  I was right there with you, man. That was awesome.

• 1:31:45Speaker 03

  Did you get goosebumps or did you cry? What was your physical response?

• 1:31:48Speaker 05

  Yeah, all those things, sure.

• 1:31:49Speaker 03

  Okay, cool.

• 1:31:50Cara Santa Maria?

  Oh, yeah, it happens to me most of the time when I enjoy, especially like live classical or choral music. I get like goosebumps and sometimes I get like that burning sensation behind my eyes or like a lump in my throat. It's pretty common for me.

• 1:32:05Speaker 03

  Okay, no specific one that was like the most that you can think of?

• 1:32:09Cara Santa Maria?

  No, I mean, I was a singer growing up and I've been to like, you know, a lot of operas and a lot of just like choral performances. And of course, I grew up Mormon, so I was in church a lot. And I had a lot of really strong emotional responses to music, which I think as a child, I probably like mistaked for feeling the spirit. But, you know, quickly realized as a young teen that like, no, I just really like music.

• 1:32:33Speaker 03

  Follow up question for you, Kara. I didn't know you were a singer. Did you ever feel goosebumps or like really intense emotions while you were singing?

• 1:32:42Cara Santa Maria?

  Yes, but never when I was, well, I shouldn't say never, but I don't think often if I'm soloing, but in a chorus, when you're hearing all of the harmonies kind of around you, I have.

• 1:32:53Speaker 03

  Do you have one, Bob?

• 1:32:54Bob Novella?

  I mean, I've experienced this many times. And I'm looking at ways to describe this. And the one, this one that I've come across, a thrilling shiver is a good way to describe it with the, you know, the skin tingling, the goosebumps, a rush of pleasure. I've experienced it many times, trying, a lot of times with music, just a specific song that just really was very emotional and moving and striking. I've experienced it. I remember experiencing it when reading about some different new types of technology that was just made me so excited. One was about, I was reading about this new technology that was super high, you know, high resolution, super high bandwidth. Everything was like at the maximum.

• 1:33:39Bob Novella?

  And I remember reading the description the writer was making of it describing what he was seeing. And I remember getting that sensation from that description of the technology. And this is like decades ago, but that's the only specific.

• 1:33:53Speaker 03

  Sure.

• 1:33:53Bob Novella?

  If I wanted to get specific, I couldn't tell you one of the songs it's happened with. It's happened to many songs and I can't dredge up the name of the song. But that one specific thing I was reading did it to me once and that was very memorable.

• 1:34:04Speaker 03

  Yeah, yeah, it's definitely not an experience that is limited to music. And it can be with poetry. It can be with the visual art, political speech. It could be anything, but I'm focused on, I'm focused on music.

• 1:34:20Speaker 05

  Jesse, did you ever look into this relationship to ASMR?

• 1:34:24Speaker 03

  I don't experience that personally, so I haven't really done much research. I like to research based on my personal experience. How about you? And do you experience ASMR?

• 1:34:35Jay Novella?

  I do not experience it. I wrote about it like 10 years ago, so a lot of people think I'm an expert on it. It's like, no, I'm just writing, just wrote a frigging blog post about it. But, so I keep getting contacted by people about it. So that's basically the.

• 1:34:45Cara Santa Maria?

  I think it's, I think it's a very similar.

• 1:34:48Jay Novella?

  It's, I think it's just a.

• 1:34:48Cara Santa Maria?

  Incredibly similar.

• 1:34:49Jay Novella?

  Yeah, an extreme version of that. That can be provoked by kind of weird things like watching somebody wrap a present. You get that tingly sensation up your spine.

• 1:34:59Cara Santa Maria?

  Yeah, but a lot of ASMR is noise focused anyway, so, yeah.

• 1:35:05Speaker 03

  Okay.

• 1:35:05Cara Santa Maria?

  Things similar.

• 1:35:06Speaker 03

  Cool, yeah. So I've been doing a lot of research on this and I'm an inventor. I invented a way where I think I can elicit musical chills a lot more frequently. This is a rare experience for most people. Would you all agree that it's very rare and difficult to find a song that gives you a really intense experience?

• 1:35:30Bob Novella?

  Yeah, it's way too rare, man. I wanna, I wanna experience this every day.

• 1:35:33Cara Santa Maria?

  Not for me. I think I could evoke it on demand if I went to, you know, the opera or the symphony regularly.

• 1:35:42Speaker 05

  I don't think it's rare for me. Plus, you know, it's about 100 times more common if you pair music and images or video.

• 1:35:51Speaker 03

  Oh, absolutely. Yeah, I use YouTube for most of my experiments and I find that to be very true. However, with one huge caveat, most YouTube music videos, the ones that are produced are terrible. You have to look for the live recordings if you wanna get, you know, some legitimate emotion from the artists and not some distracting, trippy video or whatever.

• 1:36:17Bob Novella?

  Okay.

• 1:36:18Speaker 03

  Yeah, so there's a lot of research on what causes frisson or musical chills. Or just, there's not really one all encompassing term that describes really intense musical experiences because there are all sorts of different kinds. There's a researcher that I, a professor that I recently spoke to, Scott Bannister, who does a lot of these studies. And he, he's trying to find out what causes musical chills, what parts of the music do. So he would isolate a certain part of a song, like a crescendo or a guitar solo,

• 1:36:58Speaker 03

  and see, hey, do people have musical chills in this song if we remove this particular place? And he found that, yes, they would cease to have this response at these specific triggers of musical chills. So you can isolate these triggers, but the problem is, this is a very subjective problem. What triggers frisson in me is, first, it's pretty universal that surprise is an aspect of it. So if you wear out a song, your surprise goes down. Your favorite songs, maybe the first time you heard it, you got a huge dopamine hit from it.

• 1:37:42Speaker 03

  And that's why it became your favorite song. But if you wear it out, you know, you no longer get that. You can't, you can't reproduce it if you, if you know the whole song front to back. You don't, you no longer have that surprise trigger. But I think.

• 1:37:55Bob Novella?

  That's what I do, man. If I find an exceptionally amazing song, I will kill it to death.

• 1:38:00Speaker 03

  Yeah, we all do.

• 1:38:01Bob Novella?

  But like, I read it, read it, listen to it over and over. I'm like, yeah, the magic is gone.

• 1:38:06Cara Santa Maria?

  I wouldn't say that we all do because I have songs that no matter how many times I listen to On the Nature of Daylight by Max Richter, it always gives me frisson, every single time.

• 1:38:15Speaker 03

  If you give it enough of a rest, I bet if you listen to it every hour, it would cease to do that for you. No?

• 1:38:22Cara Santa Maria?

  Yeah, maybe. I don't know. I think it's about attending to it.

• 1:38:27Speaker 05

  Maybe just a freak, Kara.

• 1:38:29Speaker 03

  Maybe. I mean, some people are more sensitive than others. Kara actually did a segment on music hedonia and hedonia about a year ago. And some people just get no pleasure from music for whatever reason.

• 1:38:46Cara Santa Maria?

  So sad.

• 1:38:47Speaker 03

  Yeah, so there are many different triggers. There are different things that can kill the effect. Like, for example, if you don't have your volume turned up, if you have crappy headphones, if you're being observed. So one problem with with musical, with studies about musical chills is that people feel like they're being observed, even if they're not, even if they're isolated in a room. Their bodies respond as if they're being observed. And so musical chills are muted for them. So I'm working on an app that would, that will make it so that people can be in whatever environment they want, just listen to their stuff.

• 1:39:29Speaker 03

  And it records your, there's, there are two ways that I, I record intense emotional reactions. Heart rate, which is surprisingly accurate. Not, these both have problems, but also EDA, which measures the conductance of your skin. So if you are having an intense emotional reaction, freeze on, it's very good at detecting. Your skin conductance goes up. And so it can detect, hey, this person is having an intense emotion right now. Traditionally, it's been used to detect stress, but it can also be used to detect strong emotions. And lots of researchers use it for that. But I'm trying to get it on a mass market smartwatch.

• 1:40:14Speaker 03

  Android, a lot of Android watches already have these two sensors. Obviously, all smartwatches have a heart rate sensor these days. But I'm working on some software that can measure this. And it can tell you, hey, you're having a, an intense reaction here. And the goal of that is, do you guys know what? Anybody familiar with the term collaborative filtering?

• 1:40:40Speaker 05

  No.

• 1:40:41Speaker 03

  So it's, it's actually pretty commonly used in music recommendation algorithms. Collaborative filtering basically sees, hey, you thumbs up all these different songs. Who else thumbs up all these same songs? And so it puts you in a bucket with that other person. And it says, hey, these people are similar. If they all thumbs up the same stuff, what are the odds that if they're given each other's songs, they're going to thumbs up them too? And it's, it's pretty effective. There are some studies on that. Collaborative filtering is pretty effective at predicting what people will like. Problem is thumbs up is ambiguous. You, you might just vaguely enjoy something.

• 1:41:25Speaker 03

  You, you, there's no, this gave me goosebumps option. And that's where the, the sensors come in. If we do, if I do collaborative filtering on people's reactions to music, it's their bodies telling, it's their bodies telling the algorithm, hey, I had this, this reaction to this specific performance. And someone somewhere else in the world had the same reaction to say 10 of the same songs. They share their songs. What are the odds that the, that the, the new person, the person's new song pushed from that other person that matched with collaborative filtering?

• 1:42:06Speaker 03

  That, and what are the odds that, that they also have a very strong emotional reaction? So that's, that's the inventor part of it. I want to make it so that you can find new music. Cause I found it very difficult to find new music that gives me free song. And I really enjoy the experience. And, you know, like Kara, Kara has mentioned that she, she has her group of songs and she enjoys, she enjoys those. She doesn't need to branch out to new stuff. But I have maybe 100 now. And I've been really rapidly searching for the last couple of years. And that's not enough for me because I'll wear one out. And then I've got nothing.

• 1:42:51Speaker 03

  And I, I don't like to bother with background music and all that.

• 1:42:55Jay Novella?

  You can, this could help you find music that could, could potentially give you that experience?

• 1:43:01Speaker 03

  Yeah. So the idea is you use collaborative filtering. And if you had intense experiences to a lot of different songs, you get matched up with a bucket of people that had the same intense experiences in the same songs. And it, it infers that, hey, you're going to have the same, you're going to have intense experiences to whatever other songs they have that you don't have.

• 1:43:22Jay Novella?

  So what data do you have that this is effective?

• 1:43:25Speaker 03

  Very good question, and you're right to be skeptical. I have a very small sample size, me and some of my acquaintances, that I've, I've measured, measured responses and I've got success rates. Important caveat, it's a very small sample size, but I've gotten success rates as high as 60%. If I listen to somebody's new song that I've never, never heard before, if we matched, I had about a 60% chance of, of feeling an intense emotional reaction to their music that I've never encountered before. That is personal experience, so it needs to be discounted. And for that reason, I'm, I'm working on a study

• 1:44:07Speaker 03

  to try to prove it with the general, with a bigger sample size, like 40 or 50 participants.

• 1:44:13Speaker 05

  And how much electricity does your app use?

• 1:44:18Speaker 03

  Well, the sensors use a lot of electricity. So you've got to have a good, good watch with, with a modern battery in it, or else your watch is going to die.

• 1:44:29Jay Novella?

  Okay. Well, I, I, this is one of those things where, wouldn't you just love 10,000 people to use the app and have that data pour in, you know?

• 1:44:38Speaker 05

  Well, that's what companies do, right? That's why they give it for free. And then that's because you're the data they're gathering, you know, for the initial release, right? It's all, so that's what you need. You just sort of give it away for free to a certain number of people. But the small print says, we're gathering data on you to show that this actually works.

• 1:44:57Speaker 03

  Yeah. And I mean, eventually I, I hope to make money off of it. I have a patent on the idea. But, you know, for now, I, I just want to make a study that proves it works because I have my cognitive bias. I think it works very strongly, but I need to prove that.

• 1:45:14Jay Novella?

  Right. All right, cool. All right, Jay, it's who's that noisy time. All right, guys, last week I played this noisy.

• 1:45:24Jay Novella?

  I, I only had two people write in with guesses. And I was expecting this, there is usually a slump in August and September that I've noticed. I think it's like summer ending back to school. And, but anyway, I got two people and they, and they both are incredibly intelligent people. You guys want to guess?

• 1:45:45Speaker 05

  So it sounds like there's birds tweeting in the background, but that's not the noise you're going for.

• 1:45:50Jay Novella?

  It's the, the noise that's in front.

• 1:45:51Speaker 05

  Then there's the roaring, like I'll call a roaring noise up front, but it's.

• 1:45:56Cara Santa Maria?

  Yeah, almost like gas.

• 1:45:57Speaker 05

  Yeah, but yeah, that is just like a throaty gas expulsion kind of noise. Could be an animal, doesn't have to be an animal.

• 1:46:04Cara Santa Maria?

  Could be an animal, could be a literal, like propane hose or something.

• 1:46:09Jay Novella?

  I thought the same thing, Kara. I thought it was like a fire kind of noise to it in a way.

• 1:46:14Speaker 03

  I had to guess, I'd say maybe a wild boar or something.

• 1:46:18Jay Novella?

  Well, I had, Michael Blaney wrote in. He said, his guess is he thought it was a hyena. And I thought that was an interesting guess. I mean, I've heard the noises that hyenas make. And I tried to find other noises because animals make more than the noise that you hear. You know, typically, you know, animals don't just make one noise. They have a noise that it becomes associated with them. So I thought maybe hyenas make some other noises. So I did listen, but I didn't find anything that got close to that. And then the second guesser is Frank Esso. And he said, I think this is someone dragging a heavy wooden object across concrete. Maybe a large crate. Or it's a sea creature that will haunt Kara tonight.

• 1:47:01Jay Novella?

  So yeah, so that, I thought that was very creative. You know, because I didn't, when I listened back to it, I'm like, okay, maybe that could be something being dragged. But this one was kind of hard because it's an unexpected noise coming from an unexpected source. So there was no winner. And this is actually a bird, Steve.

• 1:47:19Speaker 05

  It is.

• 1:47:20Jay Novella?

  And you want to take a guess at what it might be?

• 1:47:22Speaker 05

  A, what are those Australian dinosaur birds?

• 1:47:26Cara Santa Maria?

  Cassowary?

• 1:47:26Speaker 05

  Cassowary.

• 1:47:27Jay Novella?

  No, this is a Australian bustard, also known.

• 1:47:32Cara Santa Maria?

  Bustard.

• 1:47:32Jay Novella?

  A bustard or a bush turkey. See, I can't even pronounce words the correct way when they're spelled the correct way. It is, I was thinking, it can't just be busted. It's probably bustard, you know? I'm like, whatever.

• 1:47:44Speaker 05

  I got the Australian part, right?

• 1:47:46Cara Santa Maria?

  Yeah, you did. That was good.

• 1:47:47Jay Novella?

  Also known as a bush turkey. So this animal, they're, they are the heaviest flying landbird native to Australia. They're famous for that roar, the dinosaur like booming. This is a mating call. They have a nomadic lifestyle. They're found in the grasslands and savanna woodlands and agricultural plains that are throughout the mainland Australia and in southern New Guinea as well. So listen again. This is, you know, this is an animal, a relatively small animal making this noise.

• 1:48:22Jay Novella?

  Definitely not.

• 1:48:23Speaker 05

  Geez, man, sounds like a dinosaur, dude.

• 1:48:24Jay Novella?

  Now I was saying, like, imagine you're walking, you're out in the bush, right? Like, because they're saying, like, it was, they come from the agricultural plains throughout the mainland of Australia. So you're walking around and 20 feet away, you hear that. You're not going to think it's some, you know, turkey kind of bird bopping around, right?

• 1:48:43Speaker 05

  And I think that's probably the point of it, right?

• 1:48:45Jay Novella?

  I guess so. Yeah, that could be like it's anti, don't like cook me for some holiday sound, right?

• 1:48:51Speaker 05

  But he said it's a mating call, though.

• 1:48:52Jay Novella?

  Yeah, it's a mating call. Well, they might be attracted to you, Steve. That's the thing. Then you should run, I guess. All right. So I have a new noisy for you guys this week. This noisy was sent in by a listener named Anton Fomenko. And here it is.

• 1:49:21Jay Novella?

  All right, I played it in its entirety. And I'm going to give you a clue because I think this one is difficult. Because it's possible that there are people who would recognize that as it is. Like, they'd hear that and go, oh, okay, I know exactly what that is.

• 1:49:36Speaker 05

  Because of the progression of the noise.

• 1:49:38Jay Novella?

  Right, because there's a pattern of some kind in there. There's a detectable thing going on in there that certain people would be able to recognize. That's my clue. And again, I repeat, it's hard. But send me in some fun guesses, because there is a million different things it actually could be. That is such a standard, it's almost like white noise. It's such a common type of sound. So I'd be curious to see what else makes that type of noise out there. So email me if you heard something cool, or you think you know this week's who's that noisy. You can email me at [email protected].

• 1:50:12Speaker 05

  Thank you, Jay. One quick email. This comes from Yasmina, who, after writing some nice stuff, says, "We were both very impressed and not surprised with how intelligent your daughter is." Well, thank you, Yasmina. "I am a new mother and thinking about how to educate my daughter is a top concern of mine. Of course, I plan to raise her as a skeptic. I was wondering what your thoughts are regarding public school versus private school, and if you feel there is a difference in education long-term success. I imagine if you are a motivated parent at home, you can certainly overcome some of the deficiencies. But I'm curious if you have come across any data regarding this, and if you are comfortable sharing, what was your approach when making these decisions for your own children?" All right, so what do you guys think out of the gate?

• 1:50:57Speaker 05

  Any difference between public school and private school? Let's just say in the US, obviously it'd be different for every country.

• 1:51:04Bob Novella?

  Well, for one, the amount of services a public school would have at their command compared to a private school.

• 1:51:10Cara Santa Maria?

  Well, I think that this is, whoa, such a loaded question. You're right, sometimes a private school is going to have more resources allocated to it, but there's also going to be less oversight, like governmental oversight, fewer regulations about curriculum. And so a lot of private schools are going to have agendas.

• 1:51:30Speaker 03

  Yeah, I went to a private school. It was a religious one. I learned all about how evolution is wrong and all sorts of other fun stuff.

• 1:51:41Bob Novella?

  Now, wait, just to be clear, when I was talking about services, I was referring to services that a public school would have, that private schools wouldn't need to have.

• 1:51:52Speaker 05

  They have way more services for special needs kids, absolutely.

• 1:51:55Cara Santa Maria?

  Absolutely. And just for kids in general, if you're going to a public school, you're going to have extracurriculars, you're going to have, I'm assuming, a more robust and well-rounded education, simply because there are so many more options available.

• 1:52:09Speaker 05

  So, the short answer is, we don't know. Because it's very hard to compare objectively, in a controlled fashion, private and public school. What differences there are usually disappear when you control for socioeconomic status. But there are a couple of studies where you were able to compare with the same socioeconomic status, and that is cities that used a voucher system. So if you have a voucher lottery, where a certain percentage of the population of the students can go to a private school on the state, and everyone's kind of demographically equal to begin with,

• 1:52:51Speaker 05

  that's as close as you can come to a controlled comparison, right? And again, you could ask a lot of different questions about what's quote-unquote "better", but if you're just looking at performance on tests, on standardized tests, because that's something you could measure. What happens when public school students go to, say, a private high school, is in the first year, their math scores drop. They actually do worse in math than their public school peers. But then they catch up by the end of high school, and there's no really meaningful difference by the end, in terms of performance on standardized tests.

• 1:53:35Speaker 05

  Public schools have more standardized teaching of math and reading, and at that grade level, they just do a better job on average than private schools do. But students generally catch up by the end. There doesn't appear to be any long-term difference either way. So basically, if you're talking about outcomes that we could measure, again, just in general, not a big difference. Slight hit on math for the private schools. But of course, this is all going to be about the specific private school you're going to send your kid to.

• 1:54:09Cara Santa Maria?

  Exactly. And the specific public school system you live in.

• 1:54:12Speaker 05

  And what public school system you live in, absolutely. So you can't really generalize this everywhere, even everywhere in the US, even state by state. You've got to know what your options are. I would say the big picture is, it doesn't really matter that much. And if you look at overall, like what predicts good academic outcomes, it's all about the support they get at home. And so I wouldn't wring your hands about that decision, so much as have lots of books in the home, be obviously very supportive, be very encouraging, read to your kids, talk with them, engage with them on interesting questions.

• 1:54:53Speaker 05

  Don't just give them answers, actually engage with the thinking process about how we know things, et cetera, et cetera. That will do way more than whatever decision you make about where to send them to school. All things being equal, obviously, if you have, if you live in a place that has really underfunded, terrible public schools, and there's an awesome private school that you can afford, that may be a better option for you. Or you may have a great public school, and there's really nothing to be gained by going to private school. That just depends on where you are and what your options you have available.

• 1:55:26Cara Santa Maria?

  And really, like, figure out if the public school is run by a religious, or sorry, the private school is run by a religious organization.

• 1:55:32Speaker 05

  Absolutely. All right, guys, it's time for Science or Fiction?

• 1:55:40Speaker 00

  It's time for Science or Fiction?

• 1:55:50Speaker 05

  Each week, I come up with three science news items, or facts, two real and one fake, and then I challenge my panel of skeptics to tell me which one is the fake. Three regular news items, you guys ready?

• 1:56:00Jay Novella?

  Yep.

• 1:56:00Speaker 05

  All right, here we go. Item number one, a new study concludes that Venus likely had a large moon early in its life, which was likely destroyed by a later collision. All right, number two, researchers are now in the home testing phase of a robotic arm fused to the bone with muscle and nerve integration, allowing for natural control and sensory feedback, a truly bionic limb. And item number three, a new analysis of human ancestor fossils finds that Australopithecus species remained significantly arboreal in their locomotion, while Homo species were fully terrestrial. Bob, you go first.

• 1:56:38Bob Novella?

  All right, Venus having a large moon. Yeah, I'm trying to think of why to rule that out. Nothing's leaping to mind. How would they discover that out? Crap, man. I don't want to spend forever on this, so let's go to the second one here. All right, so this home testing phase of, I mean, wow.

• 1:56:58Speaker 05

  So just to clarify what this means, the guy has the arm, it's attached to his body, he's using it, and he's at home outside the lab, just using it day-to-day.

• 1:57:06Bob Novella?

  So that seems a little bit too premature at this point. I mean, first off, it's permanently attached to his bone, and then it's also integrated with muscle and nerves? That's, I mean, damn, man, I'm not tracking this nearly as closely as you are, Steve, but that seems a little bit premature, especially, you know, the guy's like on his own at home with this. So that's dramatic. I mean, that really is next level, bionic limb. All right, so let's see what this third one was. Australopithecus was significantly arboreal. Okay, so they do a lot of brachiation, Steve, is that what you're saying?

• 1:57:44Speaker 05

  Not necessarily brachiation per se, but they were in the trees.

• 1:57:48Bob Novella?

  Okay. Significantly, so what does that mean, significantly? Significantly could be, you know, 30% sounds significant, based on what I was thinking. I think they would spend very little time in the trees. But what was their environment like? Maybe it was a lot safer in the damn trees. So many questions, Steve, goddamn it. I'm going to have to meta-game this a little bit. I think you might want us to pick, I think you might assume that the robotic arm one might be like, whoa, that's too soon. So let's go with Australopithecus fiction.

• 1:58:24Speaker 05

  Okay, Jay.

• 1:58:25Jay Novella?

  Okay, so the first one, a new study about, it concludes that Venus... Likely had a large moon early in its life, which was likely destroyed by a later collision. So I find it interesting that they could do a study that would be able to tell that Venus had a moon. Maybe from its orbit around the sun. Maybe there's a telltale sign in that. I don't know. And then how would they surmise? Or why would they say it was most likely caused by a collision? How could they possibly know that? I mean, on its face, it's, it's of course plausible. Move on to the second one. Researchers are testing in the home. This robotic arm, it's fused to the bone, Steve, and the muscle. Oh, it's, it's fused to the bone with muscle.

• 1:59:10Jay Novella?

  And it has nerve integration. And they can, it has natural control and sensory feedback. I mean, come on, it sounds way too, way too technologically forward than where we, where we are right now. So that's where I am on that one. And the last one, a new analysis of human ancestor fossil finds Australopithecus species remains significantly tree walkers. Right, these guys are saying it lived in the trees and they moved in the trees. Wow. I don't know. I don't know. I mean, I would think that they would, you know, could use the trees for protection and for like hunting and like to see around themselves and stuff. But to have them completely live in the tree, like they don't have a tail anymore.

• 1:59:49Speaker 05

  It doesn't say completely. That says significantly.

• 1:59:52Jay Novella?

  But that's such a vague word then, Steve. Okay, so significantly meaning like half their time?

• 1:59:58Speaker 05

  Yeah, is that a question or a statement?

• 2:00:00Bob Novella?

  It's a question.

• 2:00:01Jay Novella?

  Okay, so some random unknown amount of time in the trees. Well, now that turns it from like a, a no to me to a maybe, because you know, they could be up there for a half an hour. And that's a lot more than not at all. I mean, were they even living in places where the trees were close enough for them to, to bop around from one tree to the next? Why am I thinking that these guys were more plains dwellers? So many questions.

• 2:00:22Bob Novella?

  Savannah, baby.

• 2:00:23Jay Novella?

  All right, thanks, Bob. I guess I gotta go with the robot arm. I mean, how could that possibly be true? I guess that one's gotta be the fiction.

• 2:00:30Speaker 05

  Okay, Kara?

• 2:00:32Cara Santa Maria?

  No idea about Venus, but sure, why not? So for me, it's between the robot arm and the analysis of human ancestor fossils. Okay, so it's in the home testing phase of this robotic arm. And I guess what is kind of surprising is that it's fused and integrated with both muscle and nerve. But I think, yeah, I think that that's, I think we've been headed towards this for quite some time. And we've had lots of incremental changes. And we've probably had things that are fused to bone. We've probably had things that are, I know we've had things that are nerve integrated. We've had things that are muscle integrated. Why not combine all three? It doesn't say it works. It just says it's being tested. The one that is bothering me and for a different reason, actually, than, than Jay pointed out, and I think even Bob,

• 2:01:18Cara Santa Maria?

  is the part about the Homo species, not the part about Australopithecus. Cause you're saying both, right? So the analysis finds that Australopithecus was significantly arboreal, but Homo, but Homo species were fully, you're saying the word fully terrestrial.

• 2:01:35Speaker 05

  Yeah, as far as could be told with whatever methodology was used in this study.

• 2:01:40Cara Santa Maria?

  I thought that there was a, at least one, if not a couple of Homo species that predated Homo erectus.

• 2:01:47Speaker 05

  Oh, yeah.

• 2:01:48Cara Santa Maria?

  Well, erectus was erect. And they named it for that.

• 2:01:53Jay Novella?

  Yeah, see, they named it for that.

• 2:01:55Cara Santa Maria?

  So wouldn't there be things earlier than the standing up Homo species? That wouldn't have been standing up? And if it wasn't standing up, why? I don't know. I just think that there's probably older Homo species that weren't fully terrestrial, that were still partially arboreal. So I think that one's the fiction for that reason.

• 2:02:24Speaker 05

  Okay, and Jesse?

• 2:02:25Speaker 03

  Yeah, I don't have a problem with the Venus one. Planets collide with moons all the time. Curious what the evidence is, like a crater or something? But I don't have a problem with it. And the robotic arm, I think is cool. And I think their experiments going on all around the world. If it's reckless, it might be in China because they don't have as much regulation. So I don't really have a problem with that one either. And I was gonna say the same thing that Kara said. I think that Homo species probably, even some of the later ones probably spent a lot of time in the trees for safety. So somebody else had mentioned safety.

• 2:03:08Speaker 03

  So I'm gonna go with item three or the Homo species.

• 2:03:13Jay Novella?

  Got it. All right, so you all agree on the first one. So we'll start there. A new study concludes that Venus likely had a large moon early in its life, which was likely destroyed by a later collision. You guys all think this one is science. And this one is the fiction.

• 2:03:28Bob Novella?

  Whoa.

• 2:03:29Cara Santa Maria?

  Oh, no.

• 2:03:31Speaker 05

  This is a very cool study, though. But it didn't conclude that has no idea whether Venus had a moon or not. And as you guys all kind of all brought up the question, how would they know? Well, they, I don't know that they could know. There's no evidence that Venus had a moon or, or what happened to it. It's all, but there, there has been speculation like if Venus had a moon, then what would have happened to it? Why doesn't it have a moon today? And one theory is that, well, you know, then something would have probably had to collide with it. Maybe that's also what screwed around with Venus's axis of rotation. But so they were looking, Bob, you're very, find this very interesting,

• 2:04:12Speaker 05

  at the tidal forces that between Venus and its moon. And they looked at all kinds of initial parameters. How fast were each spinning? How fast, how big was the moon? How far away was it? And what's interesting, so you know, with our system, right? The earth moon system. The earth is slowing down and the moon's getting pushed away. Right?

• 2:04:35Bob Novella?

  Title, title breaking. Yes, it's fascinating process.

• 2:04:38Speaker 05

  Yes, so Bob, what would, does that always happen though? If you have a moon revolving around a planet?

• 2:04:43Bob Novella?

  I mean, no, I mean, I don't think it necessarily happens, but wait.

• 2:04:47Speaker 05

  All right, this is what they found. That there's a certain critical distance. Beyond that distance, tidal forces would push the moon away. Within that distance, the

• 2:05:00Bob Novella?

  It breaks up the planet.

• 2:05:00Speaker 05

  The tidal, tidal forces will lure the moon in. So the moon would fall in toward the planet.

• 2:05:07Bob Novella?

  Until it breaks up.

• 2:05:08Speaker 05

  Until it reaches the Roche limit and breaks up.

• 2:05:11Bob Novella?

  Right.

• 2:05:11Speaker 05

  So however, that's, that's the synchronous radius. However, that synchronous radius moves outward as the planet slows down. And so at some point, right? So if the, if the moon's outside the synchronous radius, it gets pushed away. If it's inside the synchronous radius, it gets, it spirals into the planet. But the synchronous radius itself moves outward as the planet slows down. So the question is, in what scenarios would the, would a moon orbiting Venus get captured by, within the synchronous radius? And then start to spiral in rather than getting pushed further out?

• 2:05:50Bob Novella?

  If it has a, if it has a fast rotation, short days.

• 2:05:52Speaker 05

  Yeah, so it all depends on, on these variables that I was talking about. But for a lot of scenarios, that was what would happen. Cause Venus is so rotating so slowly, basically. Right? So anyway, so it's what the, the reason for the study was to say, if Venus did have a moon, would, would a later catastrophe be necessary to explain why it's not there anymore? And the answer is no. That probably that moon would have been destroyed. Depending on the variables, between 0.03 and 1.7 giga years, billion years. So and even in the range of parameters they looked at, the moon would be long gone by now. That was the purpose of the study.

• 2:06:37Speaker 05

  But they have no idea if it actually had a moon or not.

• 2:06:40Bob Novella?

  It's one of the two planets that doesn't have one. So it's kind of an anomaly. Yeah.

• 2:06:44Speaker 05

  All right. Item number two, researchers are now in the home testing phase of a robotic arm fused to the bone with muscle and nerve integration, allowing for natural control and sensory feedback, a truly bionic limb. This is very cool science.

• 2:06:57Bob Novella?

  Yeah.

• 2:06:57Speaker 05

  Kara's right, all these pieces were already in play. This is just really the first one to bring them all together. And actually, it's the first one in the US. A, a European patient already had this three years ago. But that was just a hand.

• 2:07:11Bob Novella?

  Three years ago.

• 2:07:11Speaker 05

  This is above the elbow. This is a person who had an industrial, you know, work accident. They had their left arm amputated above the elbow. In July, they had this surgery, which fused the titanium base to their bone and also integrated the electrodes to their muscles and to their nerves. And now they're, they had it recently, actually, they had, then they tested the arm. It's cool. You could see a video like over, over the summer where they're, they're testing the arm. The arm is just plugged into the prosthesis on his arm. But it's not the arm, the robotic arm is not attached to his body. There's just a wire from his arm, the stump of his arm to the prosthesis.

• 2:07:56Speaker 05

  And they're just, and they're calibrating it and testing it to make sure that it functions. Well, now they've actually attached it and sent them home to see and to use it.

• 2:08:03Bob Novella?

  Well, what's, what's the, I mean, how does it operate? Is it, is there some sort of BMI, brain machine interface? Or is it based on like a muscular movement to make it move?

• 2:08:12Speaker 05

  Muscles and nerves. So he, when he moves the muscles that, that would move to move his arm, those little muscle movements get translated into his, his desired movement. And guess what makes that translation? AI.

• 2:08:27Speaker 03

  AI.

• 2:08:28Bob Novella?

  AI.

• 2:08:28Speaker 05

  So AI is used to translate the muscle movements into his desired limb movement. And then-

• 2:08:35Bob Novella?

  But more importantly, Steve, it's not, because I've seen these devices operate where if you operate your chest muscle, it moved the arm and it looks very, it just looks weird and unnatural. But these, you're talking about the muscles that he would be using anyway.

• 2:08:51Speaker 05

  Well, not necessarily, because you only have the muscles above the elbow. So none of the muscles below the elbow that would control your fingers.

• 2:08:59Bob Novella?

  No wrists, no fingers, yeah.

• 2:09:00Speaker 05

  So it's not the same-

• 2:09:01Bob Novella?

  He's got biceps.

• 2:09:01Speaker 05

  It's not the same muscles that he would have been using to move the hand, right? But it is muscles that he would be using to use the arm. And he had to learn-

• 2:09:09Bob Novella?

  Okay.

• 2:09:09Speaker 05

  -how to control the hand with those muscles. And the AI had to learn what he intends to do based upon the pattern of muscle contraction. And there's also sensory feedback. So there are sensors on the robotic limb that feed directly to the nerves, which they basically, when they had to take the nerves previously and connect them to, again, muscles in the remaining stump to keep them alive and basically keep them functioning. And now they attach sensors to them. So he has sensory feedback. He has motor control. It's all fused to his bone. And it's like, pretty much all the elements are there.

• 2:09:51Cara Santa Maria?

  And then he just has to like, learn to use it.

• 2:09:53Speaker 05

  Yeah, which is already, he's already been doing, right? Because again, even before they, he was healed enough to physically attach it. Because you could still connect it with a wire still. He has the same control he has over it now. It's just now physically attached to the base, which is fused to his bone.

• 2:10:09Bob Novella?

  So I assume AI sped up the training process?

• 2:10:11Speaker 05

  Oh, yeah.

• 2:10:13Cara Santa Maria?

  Do you think as a neurologist, though, that having a brain-machine interface along with this system would make that training pretty seamless?

• 2:10:22Speaker 05

  So that's a great question.

• 2:10:23Cara Santa Maria?

  Because then you're thinking about moving your hand.

• 2:10:25Speaker 05

  I don't know the answer to that, because it's, I would say, not necessarily. It's still, you're thinking, I want to close my fist. And that signal has to get to the robotic limb somehow. Whether it goes from, through electrodes on the surface, the scalp of your brain, or through the muscles, it may not really matter.

• 2:10:44Cara Santa Maria?

  But that's the difference, right? Like, if you're thinking, I want to close my hand, and you still have mapping there-

• 2:10:50Speaker 05

  Yeah.

• 2:10:50Cara Santa Maria?

  -from your hand, that's very different than, I want to flex the part of my stump.

• 2:10:56Speaker 05

  Yeah, but here's the thing, Kara, you're right. Except the brain appears to be very plastic when it comes to that.

• 2:11:02Cara Santa Maria?

  Yeah, it's going to remap it over time.

• 2:11:03Speaker 05

  It will remap. And then, once that happens-

• 2:11:05Cara Santa Maria?

  And that's why I asked.

• 2:11:06Speaker 05

  It'll be fully natural.

• 2:11:08Cara Santa Maria?

  Yeah, I wasn't asking, will this, would that be better overall? What I specifically asked was, would it be faster to train that way?

• 2:11:15Speaker 05

  Oh, I don't know, maybe.

• 2:11:16Cara Santa Maria?

  Yeah, because you already have the architecture.

• 2:11:18Speaker 05

  Yeah. I think it depends on the exact nature of the hookup.

• 2:11:23Cara Santa Maria?

  And for sure, if somebody had a congenital, you know, difference, or a very early traumatic injury versus this just happened.

• 2:11:30Speaker 05

  All right, let's go on. Number three, a new analysis of human ancestor fossils finds that Australopithecus species remained significantly arboreal in their locomotion while Homo species were fully terrestrial. This has been a longstanding debate in the community about, when did we stop, basically, moving through the trees? This is, the evolution of human bipedalism was taking place at a time when the forests of Africa were giving way to the savanna, right? So we just sort of adapted to this newly forming savanna. But which species were still in the trees and which ones were no longer in the trees is an open debate. And we use morphology to answer that question. But this study used a different piece of information that may be better. And any guesses on what that might be?

• 2:12:15Speaker 05

  And they said, what does it mean to be significant? And this answers that question very specifically.

• 2:12:20Cara Santa Maria?

  So you said it's not morphology?

• 2:12:22Speaker 05

  Not morphology.

• 2:12:24Speaker 03

  My guess would be injuries from falls from trees.

• 2:12:28Speaker 05

  Nope. Good guess, very good guess.

• 2:12:30Cara Santa Maria?

  Physiology. So if not anatomy, then something, some indication of physiology?

• 2:12:34Speaker 05

  Yeah. But you can get from the bones.

• 2:12:37Cara Santa Maria?

  So how they were using-

• 2:12:37Speaker 05

  That you can get from the bones.

• 2:12:38Cara Santa Maria?

  Like, strength or reflection or something?

• 2:12:41Speaker 05

  Oh, yeah.

• 2:12:42Cara Santa Maria?

  Yeah, there you go.

• 2:12:43Speaker 05

  Bone density because-

• 2:12:43Bob Novella?

  If you still have very, you know, very thick-boneed arms and shoulders.

• 2:12:47Speaker 05

  Exactly. So you could look at humans and look at the bone density in the arms versus the legs. You could look at modern apes, extinct apes, and look at their bone density in the arms versus legs. And then you can compare that to Australopithecus species and Homo species. So, you know, humans have greater bone density in our thighs than we do in our arm bones. Whereas apes-

• 2:13:11Bob Novella?

  Speak for yourself.

• 2:13:12Speaker 05

  -have both, because they walk, but they also still use their arms to climb trees. So the Australopithecus species had arm and leg bone density similar to arboreal apes, while the Homo species had arm and leg densities similar to humans. So, meaning they weren't arboreal enough to increase their arm bone density. That's what's in by significantly in this context. Enough to alter their bone density. Whereas Australopithecus looked like they were still significantly arboreal enough to give them the arm bone density of modern apes that are still arboreal. Does that make sense?

• 2:13:50Bob Novella?

  This isn't, yes, this is a new discovery?

• 2:13:53Speaker 05

  Let me clarify that. So they, this individual study that I'm referring to actually just adds more individuals to existing research. So they had done this analysis on other species, but now they've done it on another Australopithecus and another Homo species, showing that that pattern holds out. And they're saying this may be the functional difference between Australopithecines and Homo genuses. And, which is interesting, because I talked about this not too long ago, that the guy who was saying, ah, just make them all Homo, like, just make them all one genus.

• 2:14:24Bob Novella?

  Right, right.

• 2:14:25Speaker 05

  But maybe that is no longer appropriate if we could say, ah, here is an objective functional difference between Australopithecines and the genus Homo that we could measure in fossils.

• 2:14:37Bob Novella?

  I like it.

• 2:14:38Speaker 05

  That would be great, yeah. All right, well, I swept you guys. This was a tough one this week.

• 2:14:43Bob Novella?

  Yeah, that was a tough one, boy.

• 2:14:45Speaker 05

  It was a tough one. You guys kind of had a hint, like, how could they know that Venus had a planet, a moon? Yeah, yeah, how could they know? That's a good question. All right. Jesse, what do you think, Jeff, once?

• 2:14:54Speaker 03

  Yeah, I really enjoyed it. Thanks, guys.

• 2:14:57Speaker 05

  Good, then you're a winner. Good. You're a winner, then you're a winner. All right, Bob, give us a quote.

• 2:15:03Bob Novella?

  Okay, so I went with a fun quote, more of a fun quote than I normally would have picked, because I needed a fun quote this year. So Richard Feynman, oh my God, look him up if you're not familiar with Feynman. Just an iconic, iconic scientist. He said, probably quite a while ago, "Physics is like sex. Sure, it may give some practical results, but that's not why we do it." So, yeah, it's just, yeah, haha, it's funny. But, you know, discovery and learning, just for the pure joy of it, without worrying about how you're going to apply that to, like, battery technology or spaceships, that's fine. That joy is that unbridled joy with just pure discovery with no other real uses.

• 2:15:47Bob Novella?

  You know, that's, you know, one of the essences of science and being a scientist, for sure. But of course, the practical results are also quite nice as well. So it's kind of a, I think it's kind of a dual thing.

• 2:15:59Speaker 05

  It's a win-win.

• 2:16:00Bob Novella?

  Yeah.

• 2:16:01Speaker 05

  Yeah, it could be, it's, science could be fun and sexy, but also change the world. Absolutely. All right, well, thank you all for joining me this week.

• 2:16:11Bob Novella?

  My pleasure, Steve.

• 2:16:11Cara Santa Maria?

  Sure, man. Thanks, Steve.

• 2:16:12Speaker 05

  Jesse, thanks for joining us.

• 2:16:14Bob Novella?

  Thanks, Jesse.

• 2:16:15Speaker 05

  And thanks for your support, of course. We always appreciate our patrons. And until next week, this is your Skeptic's Guide to the Universe.

• 2:16:26Speaker 05

  Skeptic's Guide to the Universe is produced by SGU Productions, dedicated to promoting science and critical thinking. For more information, visit us at theskepticsguide.org. Send your questions to [email protected]. And if you would like to support the show and all the work that we do, go to patreon.com/skepticsguide and consider becoming a patron and becoming part of the SGU community. Our listeners and supporters are what make SGU possible.