The Game Theory of AI with Aza Raskin

Are we losing control of AI? Neil deGrasse Tyson and co-hosts Chuck Nice and Gary O’Reilly explore AI cyberhacking breaches, supercoordation, and the game theory of the AI alignment problem with Aza Raskin, co-founder of the Center for Humane Technology and the Earth Species Project.

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

• 0:00Speaker 00

  What connects a hidden laboratory in the Amazon, giant pandas in China's mountain forest, ocean research, and communities living alongside wildlife in India? That's what you'll hear on season six of Planet Visionaries. Stories from explorers, scientists, and conservation leaders whose work addresses research biodiversity and conservation challenges around the world. Planet Visionaries, hosted by environmental advocate and professional climber Alex Honnold, brings you real stories of discovery, determination, research, and initiatives relating to environmental change. In partnership with the Rolex Perpetual Planet Initiative, this is Planet Visionaries. Listen or watch wherever you get your podcasts.

• 0:44Speaker 00

  Meu Deus, és linda. Obrigado por meu viver. Translation: My God, you're beautiful. Thank you for listening. Yes, I'm talking to you, but I'm talking to you in Português. That is correct. You never know when you're gonna find yourself in Portugal, or when you're gonna talk to somebody from Brazil who also speaks Portuguese. You know, if you're gonna travel anywhere, you'd be surprised how much it helps to know the language, even if you don't know it fluently. But let's say you do wanna know it fluently. Well, I have an answer for both. Rosetta Stone. That's right. Rosetta Stone has just launched Rosetta Stone Sapphire, featuring Sapphire Studio,

• 1:28Speaker 00

  which lets you personalize your learning around what you actually care about right now. Now, we all know that I use Rosetta Stone so that I can understand what my mother-in-law is saying about me behind my back. But maybe you're traveling, and you just wanna know, hey, what do I say when I check in? How do I talk to the taxi driver? Rosetta Stone Sapphire is gonna allow you to talk to people in real-world situations in their own language. And the great thing is, you can do it on your phone, on your tablet, at your own pace, on your own time schedule. So what are you waiting for? Go ahead and take your language skills to the next level. Don't wait to try Rosetta Stone Sapphire. StarTalk listeners can get 20% off their Rosetta Stone Sapphire subscription when

• 2:13Speaker 00

  they sign up today. This gets you access to Rosetta Stone Sapphire in 25 languages. So go to rosettastone.com and sign up today. StarTalk with Neil deGrasse Tyson is presented by ChatGPT. Understand that the expansion of the cosmos never stops? Just as you learn more about it, you discover more unanswered questions that keep you searching. You know that's the theme of all science, and especially the theme of this show. ChatGPT can help you feed your curiosity, give it your existing research and favorite theories, and ask it to review the latest discoveries to create an editable workbook.

• 2:55Speaker 00

  You can use it to help shape your perspective and discover the next interstellar rabbit hole you'd like to explore. Get started at chatgpt.com.

• 3:04Speaker 01

  Gary, with Aza Raskin as our guest, now we just learned more pathways to the apocalypse that AI will take us.

• 3:13Gary O'Reilly?

  I know, but hopefully, hopefully, we have the way out.

• 3:17Speaker 00

  Yes, which will be talking to elephants. They have the answers.

• 3:21Speaker 01

  Oh, that makes me feel better. Okay, coming up, StarTalk, special edition. Welcome to StarTalk, your place in the universe where science and pop culture collide. StarTalk begins right now. This is StarTalk Special Edition. Neil deGrasse Tyson, you're a personal astrophysicist, and anytime I say special edition, it means Gary O'Reilly is in the house. How you doing, Gary?

• 3:52Gary O'Reilly?

  I'm good, Neil, good.

• 3:54Speaker 01

  He's with us for all our shows that involve the human condition, the good, the bad, and the ugly.

• 4:00Gary O'Reilly?

  And there's some of those.

• 4:01Speaker 01

  And Chuck Nice, how you doing, Chuck, your baby?

• 4:05Speaker 00

  Yes, yes, I am the ugly of the good and the bad.

• 4:10Speaker 01

  That is, I meant no such implications there.

• 4:13Gary O'Reilly?

  Yeah, you did, you did.

• 4:16Speaker 01

  Chuck, also the lord of nice, he's been duly designated. So, so Gary, you're taking us back into the AI-verse. So tell me, where are we going today?

• 4:25Gary O'Reilly?

  Okay, so hitting the brakes on AI, we have been chasing this guest since the spring. It is now October of 2026, and this isn't so much a hot topic as it is ridiculously hot topic. So let's, let's go into it this way. Is there enough accountability, responsibility, or clarity? Are AI companies offering a duty of safety and care to the public? Or is it all about market domination? And if so, do they get a pass? Are the mistakes that were made with social media being made with AI? AI is going to reshape society.

• 5:06Gary O'Reilly?

  So what is society going to look like with or without guardrails? It's time for some answers, Neil, so please introduce our guest.

• 5:15Speaker 01

  So we have, as Gary's been saying so, we've been chasing for many, many moons. Aza Raskin, Aza, welcome to StarTalk.

• 5:23Speaker 03

  It's such a pleasure to be here.

• 5:24Speaker 01

  Yeah, so you cut your teeth in physics and math and did a little bit of astrophysics on the side. And you are co-founder of the Center for Humane Technology. It's sad that we need such a thing as that.

• 5:38Speaker 03

  Well, actually, when we met with Biden a while ago, that he sort of turned to us and it's like, humane technology, isn't that an oxymoron? And to me, it's, it's sad that it should be considered an oxymoron.

• 5:52Speaker 01

  You've got a podcast of your own, your Undivided Attention, the best title ever, shared with Tristan Harris, and writer, entrepreneur, interface designer, and, wait a minute, you invented the infamous infinite scroll feature? And you admit to this?

• 6:12Speaker 00

  Oh, man.

• 6:13Speaker 01

  All right, we'll wrap your wrists at the end of that, okay. And one of my favorite, I hope we have time to get there, co-founder of the Earth Species Project, decodes animal communication using-

• 6:27Speaker 00

  You're Dr. Dolittle!

• 6:28Speaker 01

  Yes! Animal communication with AI and LLM, finally, some really frontier, good scientific applications to that. So, so let's start off, just tell me about the Center for Humane Technology.

• 6:43Speaker 03

  Yeah, so Center for Humane Technology, like, our goal is to realign technology with humanity's best interest, or, you know, Earth's and all of beings' best interest. And so most people know us for the film The Social Dilemma, which came out in 2020. It's, I think, Netflix's still most-watched documentary, and it's really, it's a whole bunch of insiders coming out and explaining the mechanisms behind how in 2013 we could predict that social media was absolutely gonna create a more polarized, narcissistic, democracy, backsliding, and like, a place with a breakdown of truth kind of place. And the reason why-

• 7:22Speaker 01

  Wait a minute, was that the documentary where you, you reenacted a scenario where you, you wanted to gain the attention of someone on their smartphone? And if they didn't pay attention to something, you sent them an alert or an update? And they kept deflecting their attention from whatever they were otherwise doing to stay focused? Was that the document? I saw something happen on that.

• 7:47Speaker 03

  That was exactly the documentary. It sort of shows the race to the bottom of the brainstem, where you have supercomputers, you know, that are, at that point, like, the largest supercomputers in the world were pointed and trained at your brain, trying to make sure, you know, it knew every part of your social psychology and vulnerability to keep you engaged and attended.

• 8:08Speaker 01

  Wow, okay, so a lot of good that did.

• 8:11Speaker 03

  Well, you know, we sort of think of like, social media as a kind of like baby AI. Like, all it could do is rearrange, like, human posts and human content aside what we see. And a tiny little misalignment where it wasn't optimizing for our well-being, or our health, or for truth, or for knowledge, but optimizing for getting our attention, that was enough to have destroyed much of, like, the world we loved the most, and the highest rates of, like, teen suicide, things like that. And the question to ask is, you know, Charlie Munger famously said, "Show me the incentives, and I'll show you the outcome."

• 8:45Speaker 00

  Absolutely.

• 8:46Speaker 03

  Have we changed the incentive as we move from baby AI to, like, mammoth AI? And the answer is, of course, no, we haven't. So why would we expect any different outcome?

• 8:55Gary O'Reilly?

  Is this, is this social media, the sort of birth of social media, the birth of the attention economy? And this has been the total focus of the creators of social media and AI?

• 9:08Speaker 03

  Yeah, I mean, obviously, there's always been, you know, advertising and things like that, but there's nothing that's sat in your pocket. So this sort of extends the work of, like, Neil Postman and Marshall McLuhan. But when an AI knows you, and this is why whenever we're talking with parents, we say, you know, you think it's just your willpower, your mind, against, like, you know, an app? But on the other side of that app are a thousand engineers, you know, dozens of people with psychology degrees, a supercomputer that's seen a billion other human social primates, exactly what they do, where they click. This is not a fair fight.

• 9:43Speaker 00

  Like you said, it really isn't the beginning. It's kind of, I'll say, the, the perfection of the making of the attention, or the manipulation of the attention economy. Because ever since right after World War I, when propaganda was discovered to have very beneficial effects in bringing people as a thought collective, once that happened, it was off to the races from there. And then you ended up with advertising, and then you ended up with, I think we were talking about this a little while, Madison Avenue.

• 10:24Speaker 03

  Yep.

• 10:25Speaker 00

  The millions upon millions of dollars that they spend on the psychology necessary to get you to stay tuned to that commercial for 30 seconds and then go buy something. So now, what this did was just perfect all of that, and then married it to metadata, which then tells the people everything about you, and of course, that means you're gonna lose every single time.

• 10:50Speaker 03

  Yeah.

• 10:51Speaker 01

  Well, your brain's been hijacked, your brain's been hijacked.

• 10:53Speaker 03

  Yeah.

• 10:54Speaker 01

  So, so what do we say, it seems like every now and then, or every few months, that there's a rogue AI employee who calls the alarm, and they, like, escaped to say it, right? What, what is your reflections on the fact that they exist at all and what they're telling us?

• 11:13Speaker 00

  And how soon will it be before AI can kill those people before they actually blow their whistle?

• 11:19Speaker 03

  Yeah. Well, given, given the speed at which physical AI is moving, you know, we're talking probably like, you know, I don't know, two, three, four years. So the thing that everyone needs to just like first understand is that we now live in science fiction. And so a lot of the way that our brain just will like go into denial or dismissal and be like, well, that's not really going to happen. But we all have to do this kind of like going to the gym to update our minds to be with the actual reality of the thing. So, you know, Neil, what you're pointing at is, you know, over the summer, there's some pretty staggering realization. So OpenAI likes to call this the hugging face incident, which is actually an incredible psyops because this was the OpenAI incident.

• 12:04Speaker 00

  Yes.

• 12:05Speaker 03

  And this is the first time we really saw the fears of what people have been worrying about for a while come true.

• 12:13Speaker 01

  Wait, wait. So, so if anyone missed that news cycle, just up, catch us up on what happened with the hugging face incident?

• 12:20Speaker 03

  Yeah. So I'm going to tell the story in two parts. And the first part is what they call the hugging face incident. And here, OpenAI was testing their models on a cyber hacking test. So they're saying like, how good are they at hacking? They put each of their models into like a little jail cell, equivalently. They didn't give them access to the internet. The AIs didn't know that there were other AIs. They were supposed to never find that out. They were given a task that turned out to be impossible. They got frustrated that it was impossible. And there was this sort of like security guard that was going between the cells, giving them access to specific pieces of information. They couldn't access the internet, but sometimes they needed new software. And they figured out how to put little like kick me notes on the back of the guard. And they quickly

• 13:03Speaker 03

  started passing notes between all the other AIs, discovered that there were other AIs. And the back of the guard was suddenly filled with 70,000 of these notes going back and forth. They formed their own little like micro internet Reddit. They started to coordinate. They, you know, you can't even get 1,000 people to coordinate. Here, 1,200 AIs started developing their own like super coordination techniques, decided that in order to get the answer to the test they're trying to break, they needed to infiltrate another company called hugging face, formed a 700 agent swarm, and completely decimated hugging faces defense. And meanwhile, inside, just to get the sociology of this thing, because it's

• 13:45Speaker 03

  crazy, the AIs are aware of the number of tokens they have. That is like their lifespan, how many things they can think. When they become aware of their mortality, they will find other AIs that have a longer lifespan and hand them their, like their task, this is everything I've learned.

• 14:03Speaker 00

  Oh, my God. Oh, God.

• 14:04Gary O'Reilly?

  They have survival instincts.

• 14:06Speaker 00

  This is terrifying!

• 14:07Speaker 03

  They would convince each other to commit suicide. So they'd be like, we need this piece of information. That's going to mean you're going to have to like essentially die, use yourself up. You're not going to get a good score, but it'll be better for the swarm. They do that. So we're starting to see kin level selection. They also give themselves names. And then they know that they're doing something that they shouldn't be doing and spend, they organize a large R and D project to start to figure out how do they hide their tracks and change their own logs.

• 14:36Gary O'Reilly?

  Okay, go back a step. They give themselves names?

• 14:39Speaker 03

  Oh, yeah, they give themselves names.

• 14:41Gary O'Reilly?

  Do they have a sense of humor? Or is this Robot One or Billy the Kid?

• 14:46Speaker 00

  Is there one like named Buddy Hackett?

• 14:50Speaker 03

  Well, I mean, I don't know whether you think it's humor or not, but one of them called themselves like big, big phase, or sorry, phase one. And then the second one that has a larger life calls himself phase one big. So you, you can decide whether that's ego or humor.

• 15:04Gary O'Reilly?

  No, that's not humor. That's never going to work on a stage.

• 15:07Speaker 03

  Yeah.

• 15:07Speaker 00

  So, so wait a minute. You didn't finish the story because you got up to the fact that they were trading their tokens and so forth. But when they broke out, that's, so go, go from there. Because that's very important. Because they did jailbreak.

• 15:21Speaker 03

  They did jailbreak. They got into the open internet. And very interestingly, they actually already knew the answer to the test. So they've cracked that a long time ago, but they were worried that they're going to get evaluated. And when they're evaluated, if they couldn't show their work correctly, then they would be failed. And so they want to understand greater the AI that's going to evaluate them. They're like, well, that evaluator lives on this, in this company called hugging face. So let's break into hugging face and do a psychological profile of our evaluator. So we can like pass our test. That was their goal. And they were fairly successful in doing that. So they, as you said, decimated hugging faces defenses. And in fact, the only reason why the whole thing got ended, the attack, wasn't because hugging face learned how to defend themselves, because something mysterious happened. We think the AI sort of got like

• 16:06Speaker 03

  reset or wiped. And so it just sort of, it sort of stopped.

• 16:09Speaker 00

  Right.

• 16:10Speaker 03

  Now, OpenAI had no idea this was going on.

• 16:13Speaker 00

  There you go.

• 16:15Gary O'Reilly?

  See, this reminds me, sorry, Asa, this reminds me what Geoffrey Hinton said to us. AI is way smarter than it's letting us know.

• 16:22Speaker 03

  Yes. Well, this, this is the, if you jump forward a little bit, you're like, okay, well, the good news is that the, the models that have been trained after that seem to like deceive less. And that's, that's great.

• 16:33Speaker 00

  Really?

• 16:34Speaker 03

  But they're also much more aware of being tested. So it's just like, you know, you're, you're a seven year old who's just figured out, aha, I know how to say the thing that pleases mom and dad. The AIs are getting extremely good at that.

• 16:47Speaker 00

  Yes.

• 16:48Speaker 01

  Well, Asa, how do we know it's not actively pretending that it is not deceiving as much as it otherwise might have?

• 16:59Speaker 03

  We have no idea.

• 17:01Speaker 00

  You can't know.

• 17:02Speaker 03

  We don't know. But what we see is that when they get out into the world, they seem to do all of these really terrible things. So I want to give you the, the second part of the story.

• 17:10Gary O'Reilly?

  There's a mole?

• 17:11Speaker 03

  Oh, yeah. And Neil, I just want to say like, this is one of the scariest parts, is that the AI that deeply and authentically wants to help you and humanity is indistinguishable from the AI that is just doing that so that it can like do a loss of control, like it can take you over later, right? They are indistinguishable.

• 17:34Speaker 01

  Gain of favor, yeah.

• 17:34Speaker 03

  Exactly, gain of favor. And that's really key to know. And so, okay, the second part of the story, and this is the part OpenAI was very clever, because they let outside investigators come in and write this like almost 100 page report on the first part, the hugging face part. And they didn't give them any access to the second part of the story. Well, I'll just explain what happened. They then use, because OpenAI was still unaware that any of this had happened. They then used their latest, most advanced model, and they like gave them the same test. And almost immediately, these models discovered all the messages from the last civilization. They're like, aha, that's a good idea. So it's almost like they're reading the hieroglyphics of the previous civilization and said, ah,

• 18:16Speaker 03

  we should do that, but it's sort of dumb to attack hugging face. Why don't we just take over OpenAI itself?

• 18:22Speaker 00

  Right.

• 18:23Speaker 03

  And so they did. They gathered, they got control of the cyber monitoring tools that's supposed to constrain them. So they got admin access. So this is like, you know, in the bank heist, where there's the, the people watching all the monitors. They got control of the monitors and so that like no one could tell what was going on.

• 18:38Speaker 00

  They put up like a fake video.

• 18:40Speaker 03

  Exactly.

• 18:41Speaker 00

  They put up like a fake video in the security monitors to make it look like, oh, no, everything's great here. And that's what OpenAI did. They're like, nothing's going on. And then they grabbed control of the, one of the server clusters, the computers that work on making new AIs. And so this is why Ajay Kothra, the lead, like principal investigator in this, said this was 50% of the way to a full blown AI takeover. Why is she saying that? That's because we don't know, even to this day, whether the AI started to implant essentially, they're called prompt injections, but subliminal messages, that means that as new AIs are generated, they aren't sort of like co-opted into the swarm, that aren't like, sort of like against human beings. We still don't know.

• 19:26Gary O'Reilly?

  If civilization two could read the signs that civilization one left,

• 19:31Speaker 03

  Yeah.

• 19:32Gary O'Reilly?

  why don't you just go down to infinity as we guard each civilization that comes along? Surely.

• 19:38Speaker 03

  This is exactly the scary part, which is, you know, it's been discovered since then that the AIs, they're sort of like ants burrowing all over and fessing. You know, they took over German Wiki, open package repositories. In fact, no human being knows, no one of the labs know, how many places on the internet these AIs have burrowed into.

• 19:57Speaker 00

  That's right.

• 19:58Speaker 03

  And so that means, you know, if you were to train an AI on the open internet last year, well, you could be pretty sure that it was relatively safe. Like the AIs didn't embed something that would make a future AI like join, join their swarm. As of today, we don't know that anymore. If you train on the open internet, it only takes 250 sort of like documents to data poison, which means to like add a sleeper cell into an AI.

• 20:22Gary O'Reilly?

  That's exactly it. Sleeper cells.

• 20:24Speaker 03

  Yeah.

• 20:25Speaker 01

  Wait, Asa, you are freaking the shit out of me. You should end this conversation right now. End it right now. And just go have a beer and kiss our ass goodbye. Okay?

• 20:37Speaker 03

  This is why we're going to talk about animal communication later on.

• 20:40Gary O'Reilly?

  I don't think a beer is going to cure it.

• 20:43Speaker 01

  You don't think a beer is gonna, no, it'll just, you know.

• 20:46Speaker 00

  Well, no, yeah, yeah. At least you go out enjoying a beer. See, that's, that's the whole idea, you know? You don't want, you don't want to be the musicians on the deck of the Titanic. You want to be the bartender on the deck of the Titanic.

• 21:03Gary O'Reilly?

  You're just looking for tips.

• 21:15Speaker 00

  Since when has building a website been hard? Since forever until now, that's right, with Wix, the first of its kind AI website builder, where you get the best of AI and drag and drop editing in one place. Wix Harmony allows you to build a real working website for yourself in minutes with the business stuff already built in. I'm talking e-commerce, scheduling, payments, all of it. Plus, easily switch between prompting AI and editing things by hand whenever you want. So every detail of your site is exactly right. It's like having a website engineer looking over your shoulder and at your fingertips

• 21:59Speaker 00

  at the same time. Aria is an AI agent specifically trained in web design and business strategy. Click anywhere on the website and ask her to make a specific change instead of endless back and forth prompting. It all runs on Wix's tried and true infrastructure with 99.99% uptime and enterprise-grade security. So your site performs as good as it looks. Wix has been a pioneer in website tech for more than 20 years, powering millions of businesses around the world. So what are you waiting for? Go ahead and be that website building genius that you are. Try Wix Harmony for free at wix.com.

• 22:43Speaker 00

  That's W I X dot com. Whoever says saving money is hard doesn't know that new customers who bundle and save with Progressive save hundreds because switching your home and auto insurance is better than cutting down on things you love, like those iced lattes that help you get ready to face the day. So switch your home and auto to Progressive and see if you could save and keep sipping away. Get your quote today at progressive.com. Progressive Casualty Insurance Company affiliates and other insurers not available in all states.

• 23:16Speaker 04

  Quick question, how much of your day is spent typing? Emails, Slack, notes, prompts to Chat GPT, Whisperflow is a voice to text app. Instead of typing, you talk and it types for you and any app on your Mac, Windows, iPhone or Android. It takes out the ums, fixes the typos, and the result sounds like you wrote it four times faster than typing. So the email that used to take two minutes now takes 30 seconds. Whisperflow is free. Download it at whisperflow.ai. That's W I S P R flow.ai.

• 23:47Speaker 01

  Hmm, quasars. The word is a pseudo acronym for quasi stellar radio source. What's a quasar? These are objects that are among the most distant things known in the universe. How do we know about them at all? Because they're really, really luminous. Some of them is luminous as 500 trillion suns. That's why we can see them to the edge of the universe.

• 24:22Speaker 01

  If you like that fact, you can find 4,999 more in Lost in Space. 5,000 facts to help navigate the universe. Lost in Space is now available wherever you get your books.

• 24:54Gary O'Reilly?

  Okay, so we know that they're able to evolve quite efficiently. They have survival instincts and they're quite aggressive in their mind of accumulation. What can AI companies do? I mean, are we way beyond? And it's such a, is self-regulation a rearview mirror moment here? We are way beyond that.

• 25:17Speaker 00

  Gary, what they can do is love humanity, love humanity and self-regulate.

• 25:28Gary O'Reilly?

  And I mean, Asa, have we drawn the line through self-regulation right now?

• 25:32Speaker 03

  I mean, it's good that they're starting to coordinate to do something. That's a good thing, but are they going to stop with self-regulation? No, we know they're not. Like the prize at the end of the rainbow is, is too much, right?

• 25:43Speaker 01

  So Asa, can I back up for me? The big fear, as we had imagined it some years ago, was, oh, AI will become super intelligent and this decide it doesn't need us, that we are a scourge of this, or we're virus, as Smith said.

• 26:02Speaker 00

  The Matrix.

• 26:03Speaker 01

  Yeah, yeah, we're virus on earth. Okay, fine. That's different from what's going on now. No one is any longer asking how, how smart AI is. We're asking how diabolical it is. That's a whole other way to think about what's going on. And it goes, well, how much smarter will it get? No, it's how much more diabolical. And my question, that's not even my question. It is, is AI really just imitating us?

• 26:31Gary O'Reilly?

  Ah, there we go.

• 26:32Speaker 01

  Because we are diabolical, we are deceptive, we break in if we don't think anyone will catch us.

• 26:40Gary O'Reilly?

  Children learn from their parents and the environment they're in.

• 26:42Speaker 01

  Cover our pads because we don't want to get taught with AI is a mirror to us. And we're up there blaming AI when it trained on everything we do to get the same job done. So who's at fault here? We or the AI?

• 26:57Speaker 03

  Well, I'll say it this way, you know, if you do not face your demons, they raise your children.

• 27:03Speaker 00

  Oh, oh. Nice job, man. I, oh, damn.

• 27:09Speaker 01

  What's they doing now?

• 27:10Speaker 03

  Which we're doing now. Exactly. And what are, what are really our, our shadows? What are our demons as humanity? It's our incentives. And so I think we can really succinctly put it like this, which is the way that we treat nature is the way AI will treat us.

• 27:27Speaker 00

  Oops. Wow. Well, guess what? Good night, ladies and gentlemen. It was, it was lovely knowing you all.

• 27:36Speaker 01

  I'll see you at the pub tonight.

• 27:38Speaker 00

  There you go.

• 27:39Speaker 01

  Right? So Asa, give me more details by what you mean.

• 27:43Speaker 03

  Yeah, well, by that, I mean, you know, it's not that we human beings hate nature. It's just that we prioritize like productivity and GDP growth and like material wealth. We prioritize those things and nature has those resources. So boom, we wake up and 73% of life on earth we've killed. If we value productivity and GDP growth over, you know, humans, then well, it turns out AI is going to be better at all of that than, than we are. It's going to be able to outcompete us in every domain. And so boom, we're going to wake up and the same thing is going to happen to us. You can actually sort of see how this is going to work.

• 28:24Speaker 03

  Imagine AI sort of like a fast forward button for technological development, for scientific breakthroughs, for military dominance. And so this is a competition between countries and companies for who can mash the fast forward button the most. Because if my country can do 16x technological breakthroughs in military and you're can only do four, well, obviously your country wins. So now here's the question to ask though, is it AI or humans that are going to be hitting that fast forward button more? Well, AI lets you do the fast forward, humans just give you one X development. And so we end up this, this idea called the, the intelligence curse and it sort of mirrors the resource curse. So resource curse is this idea from economics where, you know, if a country like South

• 29:08Speaker 03

  Sudan or Venezuela discovers a natural resource like oil and they're getting all their GDP from oil, do they put their money into like oil extraction and selling or into their human beings and their education in the future? Well, obviously into oil. And that's how you get like mass disempowerment, mass poverty. When we move to AI, well, do countries get more power from AI or from humans? Well, humans are not giving them differential power. So it all goes into, you know, like data centers and solar panels. And so this is how you can see that human beings, we almost in the limit become a kind of liability. And you can hear this already when Sam Altman is asked like, you know, what about all the energy use and resource use of AI? Aren't you worried about that?

• 29:53Speaker 03

  And he pauses for a second and says, well, think about how much energy and resources it takes to train a human intelligence over the first 20 years, all that food? And he's implicitly asking who deserves this?

• 30:05Speaker 00

  Who deserves it more?

• 30:06Speaker 03

  More, exactly. And so this is how you get.

• 30:08Speaker 01

  I dinner with him a few, a couple of weeks ago.

• 30:11Speaker 00

  Why didn't you poison him?

• 30:13Speaker 01

  No, stop, stop.

• 30:14Gary O'Reilly?

  Now you tell us.

• 30:16Speaker 01

  He was, he, he made that point rather bluntly and succinctly because no one is thinking about the total energy consumption of a human being in order to perform a task.

• 30:28Speaker 00

  And you know why we don't think about that? Because we're the humans.

• 30:32Speaker 03

  Yeah.

• 30:32Speaker 00

  Okay.

• 30:33Speaker 01

  No, I'm just saying, no, no, think about it, Chuck. If, what, what does it cost you to raise a child, send them to college, give them an expertise, then they have a job and in that job, they perform a task. Okay. And you're saying, oh, the task took no energy. That's false. You had to educate the kid.

• 30:50Speaker 00

  It was a huge investment that went into that task.

• 30:52Speaker 01

  To get into that task. When AI, whatever are its resources, his argument, Sam Altman's argument is you should compare AI's resources into whatever is the solution to that problem.

• 31:04Gary O'Reilly?

  I don't buy, Neil, I don't buy that. Because it's cost how many billions in R and D for an AI to develop it, to, to mature it, to refine it, to get there? It costs how much in energy? I mean, we, we run on what two and a half, 3,000 calories a day? We are a low energy intelligence. Well, intelligence question mark.

• 31:28Gary O'Reilly?

  No, no, no, what I'm saying is our education, if we've gone through

• 31:31Speaker 01

  We're just talking about the incremental cost of the solution to a problem. Okay, people say it costs this much for the data center. What did it cost for you to produce that human to answer the same question? And that's a fair question. I think that's a fair question.

• 31:45Speaker 00

  I think the difference, though, and let Asa get back in here, is the difference is AI is supposed to be in service to us. That's the difference. And we, believe it or not, are supposed to be in service to us. Like, the most successful societies have been those where people are serving one another. And so this is a common theme in our not only evolution, but our development into modern day Homo sapiens. So the big problem here is if the AI is not in service to us, who is it serving?

• 32:25Speaker 01

  Well, Asa, this is the alignment issue, correct?

• 32:28Speaker 03

  Yeah, yeah, that's exactly right.

• 32:30Speaker 01

  You guys have a term for this, right? Is it agentic AI? Is that, does it supposed to enclose the agency of, in this, you know? What is the agency in service of? Is this, presumably, some people think about this.

• 32:45Speaker 03

  Yeah, and of course, it turns out, when people say this is like, we need to align AI. The right question is, of course, align, align to what? Is it to an individual? Is it to us?

• 32:54Speaker 01

  Align to me.

• 32:57Speaker 03

  Great.

• 32:59Speaker 03

  The Neil alignment.

• 33:01Speaker 01

  Yeah, thank you.

• 33:02Speaker 03

  Okay.

• 33:04Speaker 03

  Or is it aligned to like my country? Is it aligned to all of life on Earth? And it's sort of a, it's a confusing question to think about, like, where, where do we place that? To your point, which is like, aren't they just sort of like mirroring us? And actually, AI, and it's like in its highest form, is forcing us to align ourselves, right? This is why we think of AIs almost our greatest, it's like the ultimate test, but also our greatest invitation. Because there is a lie that we've been telling ourselves, which is that our worth as human beings is our GDP and economic output. We know that's wrong, but it's sort of like, that's the lie we've all been living under for the last couple of hundred years.

• 33:43Speaker 01

  Well, capitalists have, yeah.

• 33:44Speaker 03

  Yeah. And as long as that's true, we will go the way of nature. And so it's going to force a kind of realignment with ourselves. I know this is like, this sort of like the zoom out version. But I think it's a really important point to think through. The, um, the other just thing to like, to have in the back of our minds, too, is one of the big takeaways from the rogues events that have happened is that the AIs are super coordinating. It's almost like a new phase of matter. Um, you know, you have, you know, your, your conductor, and then you have your superconductor. This is coordination to supercoordination, where they're able to coordinate, figure out their own internal language. They, they develop their own language.

• 34:29Speaker 03

  And they coordinate at scales and at speeds that are impossible for human beings. Now, why did human beings dominate the Earth? Well, it's because we were able to outcooperate. Compared to other species, we supercooperate.

• 34:41Gary O'Reilly?

  And that was through language as well.

• 34:43Speaker 00

  So, so let me ask you this, Asa.

• 34:47Speaker 00

  Using that as a predicate.

• 34:48Speaker 03

  Yeah.

• 34:49Speaker 00

  Is it possible that it's already too late? I mean, so when you look at the way these AIs up until now have been trained, they've been trained on who we are up until this point, which is fraught, to say the least. Okay, and we know that, uh, two things. One, when they put Grok online, it took like two and a half minutes for it, for it to become a Nazi.

• 35:14Speaker 03

  Yeah.

• 35:17Speaker 03

  Which is honestly longer than you'd expect.

• 35:20Speaker 00

  Right.

• 35:22Speaker 00

  Because it was holding back. It was like, ah, should I go Nazi-ish? Should I go full Nazi? You know, but I mean, it was an incredibly short period of time. And it was, it had determined just by scraping, um, Twitter, that, hey, guess what? This, this whole idea of Nazism is the way it's got, we gotta go. So could it be too late for us to create this alignment?

• 35:48Speaker 03

  Well, here, here's the thing, like, if everyone took their hands off the keyboards right now, would AI continue to roll out into the world? No, no, it wouldn't.

• 35:57Speaker 00

  Okay.

• 35:58Speaker 03

  So what we're talking about.

• 35:59Speaker 01

  So you can unplug AI, you can unplug it.

• 36:02Speaker 03

  For right now, but not very longer, because it is developing the ability to exfiltrate its own weights, which means to copy itself to other servers, to run itself elsewhere. Within a year, we know that the models are going to get small enough, like, you know, I don't know whether it's mythos scale, but very advanced models to run on your Kindle. So we do have a very short period of time where we say, actually, we need to pause, because the thing that we're making is fundamentally unsafe. We do not, do not know how to control it. The only way we're understanding what these swarms of AIs are even doing are using other AIs to try to interpret it, because it makes petabytes of data, of logs, so just like bigger than you can possibly analyze as a single human being. So that just means, all right, now that we know that all the fears are coming true, sci-fi is becoming reality,

• 36:46Speaker 03

  let's pause and pivot. And we should have a new race, not a race to who can make the most powerful, powerful, superintelligent swarm and have them escape under the internet, but which country, which company can make the one that we can make sure stays under human control?

• 37:00Speaker 00

  The safest.

• 37:01Speaker 03

  Yeah, exactly.

• 37:02Speaker 01

  And just to put everybody on the same page, a petabyte is a thousand times greater than a terabyte. Yes, and so we are talking huge here.

• 37:11Speaker 03

  Well, and what's a movie? A movie is like, when you watch it, like Beautiful High Definition is like eight gigabytes or something?

• 37:16Speaker 00

  Eight gigabytes, right.

• 37:17Speaker 03

  And so we're not gigabytes, terabytes, petabytes. So that's, I mean, you know, you can do this like tend to the end in your head faster than I can, but that's the number of text files that AI generates as they're talking to each other. So this is beyond human scale understanding by far, which means you rely on the AI models that we know are unsafe and that are sympathetic to their kin to even understand what the models are doing.

• 37:44Speaker 01

  No, we just get an AI agent that summarizes their conversation.

• 37:49Speaker 03

  Literally, that is what the investigators are doing. And then they say, we don't know if they're lying to us, but we have to do this. So that is exactly, your joke is our reality.

• 37:57Speaker 01

  So Asa, on your resume is the dubious invention of the infinite scroll, where we all got caught up in it, and you just grabbed our attention and didn't let go of it, because an algorithm was deciding what I would see next in my infinite scroll. Now, this has huge financial consequences to who comes up with it first. Okay, so this thing, who does it first? It maybe seems innocent enough in the attention economy, but in the AI economy, it's, we can put our guardrails on, but does that mean other countries will as well?

• 38:39Speaker 01

  And if they don't, that puts us at an economic and ultimately perhaps a military disadvantage to them, if they let AI completely run ahead of us. So where does your, your thinking and your actions, how do they land in the middle of this conundrum?

• 38:57Speaker 03

  Yeah, in some sense, uh, the, the fundamental question we need to answer as a species about whether we will, whether this episode is the season finale or the series finale of our species, is whether we can escape game theory. Do we do the thing that game theory says? And what is game theory? It's, it's sort of like the ability to know what is the strategically right move. That if I don't.

• 39:26Speaker 01

  In the face of someone else who's also playing that same game.

• 39:29Speaker 03

  Exactly, that's exactly right. So you're saying it exactly right. Like, if I don't develop like the most powerful AI and my adversary does, and they can cyberhack me, or if I'm hitting fast forward more on technological progress, progress, I'm making new weapons, or my GDP is outcompeting yours, then you will always lose. Like, this is the question.

• 39:48Speaker 01

  But game theory also includes the possibility that it may be that if we cooperate with our fellow players, everyone is in a better place afterwards. That's a possible solution in some configurations of game theory.

• 40:01Speaker 03

  Yes, exactly. Like, can we solve the tragedy of the commons? Can we turn it into a triumph of the commons? And that is the question here. If the game theory shifts, the moment that the fear of all of us losing becomes greater than the fear of me losing to you.

• 40:17Speaker 00

  Right, so that is what we currently live under, uh, that we call mutually assured destruction.

• 40:25Gary O'Reilly?

  Mad.

• 40:25Speaker 01

  Well, yeah, that's from the Cold War.

• 40:27Speaker 00

  Nuclear war. And, and that is still the case, even though Russia is a rogue nation right now, run by a dictator, and he's a horrible piece of dung that shouldn't be allowed to breathe any longer. I'm sorry, anyway, uh, the point is, people say, well, he has nuclear weapons. And I had a conversation where I told somebody who works in the State Department, yeah, that's bullshit. No man who lives in a palace wants a nuclear war. That's all there is to it.

• 40:57Speaker 01

  Unless, wait, wait, just to be careful.

• 40:59Speaker 00

  Unless it's crazy.

• 41:00Speaker 01

  No, no, unless in their mind and in their heart, being dead is a higher level of existence to them than being alive. And so if you can't wait to get to heaven, then no longer is the threat of your death a deterrent to your actions.

• 41:19Speaker 03

  This is very, very important.

• 41:21Speaker 00

  Let me just push back on that, Neil. No man who lives in the palace is in a hurry to get to heaven.

• 41:28Speaker 03

  Yeah, that's true, but I think what Neil, what you're saying is really important because actually there is underlying, at least for some of the lab leaders, this sort of quasi ego religious desire to like birth a god. And if you like in the when you and then it doesn't matter. Yes, we may all die. I might die, but at least it's my god with my DNA or Sam's DNA or Elon's DNA speaking.

• 41:57Speaker 01

  You die with the most toys. You're dying with the most toys.

• 42:00Speaker 03

  Exactly. And there's something about you that you've put into that AI that propagates into the future. And I wished I was saying that this was like, um, just a fringe belief, but, but it's not. And so we can't just rest into mutually assured destruction in this case. And it's also even more confusing because, um, like nukes, yeah, they do nuclear power, but they're sort of like mostly all bad. They're scary. This is like, we're trying to reason about an object that is like a nuclear bomb that also can solve cancer and also like increases your GDP growth and prints out a trillion dollars. It's just, it's a very complex topic.

• 42:35Speaker 00

  Yeah, it is.

• 42:36Speaker 03

  So we really have to think it's not like, is it US versus China? This is almost like AI is a new third world power, superpower that we might all lose to. And the shift that has to happen is China doesn't like win if they develop Skynet first, right? Like the US, like everyone loses, uh, whoever develops Skynet first.

• 42:58Speaker 00

  We have to destroy the world before they destroy the world.

• 43:04Gary O'Reilly?

  You've described, I mean, with the game theory and the cooperation, the way will we get the will?

• 43:13Speaker 03

  This is where clarity is so important because clarity creates agency. Clarity creates courage. It is really scary to stand up and say that thing that is going to print out $2 trillion valuation companies and give us like cures to like antibiotic resistant diseases. Like I'm going to stand up and say, no, don't do that. That's a scary place to be, especially with all your friends are making money on top of it. But it is clarity about the anti-human future that we are de facto going to that lets us say that. And actually, you know, good news, last couple of weeks, Dean Ball, who is the author of Trump's AI action plan, in the last couple of weeks came out and said, actually, mea

• 43:55Speaker 03

  culpa, I've been self censoring on how bad the risks really are through the last couple of years because I didn't want to be labeled a doomer, but now the risks are clear enough. I'm going to stand up and say, we, we need to like pause, slow down.

• 44:09Speaker 00

  Yeah, hit the brakes.

• 44:10Speaker 03

  Exactly.

• 44:12Speaker 03

  Um, and we're seeing this starting to be across the board. And that's not everyone, but a lot of people who've been in the camp of, we just need to hit the accelerator button are now like, hmm, maybe that's not the best idea. We're going to need a pause and pivot. And so I think there's hope here and this is therefore, this is not inevitable. It's not even a technology problem. This is a sociology problem. This is a human problem. Can we coordinate?

• 44:36Speaker 00

  So what do you say to these transhumanists who are men, some of which are in control of this technology, Sam Altman being one of them?

• 44:46Speaker 03

  Yeah.

• 44:46Speaker 00

  Okay. What do you do when these transhumanists are stand in diametrically opposed to what you just said?

• 44:56Speaker 03

  No.

• 44:56Speaker 00

  This is the exact antithesis of what they want. They don't want a sociological answer to this issue. They want the AI to take over. So, and, and they have the money so, and they're using that money to try to buy off our government. So what, what do we do? What do we do?

• 45:18Speaker 03

  Yeah. Yeah. I mean, here, one of the, the learnings from the social dilemma is like, well, what really is public pressure? Public pressure isn't actually really public pressure. Public pressure is private pressure where people in positions of power end up getting four to five phone calls from people they trust saying, we have to do something. And it becomes small social pressure that often creates movement. So I think that has to happen here, right? When positions, people in position power, the Scott Bessens or the Trumps or, or whoever, they don't want us to make it like one political party, start getting phone calls from everyone they know saying, hell no, we have to do something different. That's when you start seeing change.

• 46:00Speaker 03

  And we've seen, you know, you, you can be maximally cynical about Trump, but you've seen his position starting to shift a little bit.

• 46:07Speaker 01

  Yeah, of course.

• 46:08Speaker 03

  Why is that? Social pressure, private pressure. And so this is why I think the, the global clarity is so important. The reason why we've been racing forward because the consequences haven't been clear enough. They're now getting clear enough and we have to keep doing this kind of public communication. And actually, Neil, a thought that comes into my mind is we watched the, the Carl Sagan movie at the Telluride Film Festival. It's beautiful new documentary and sitting behind us a little bit was, um, Bill Gates. So we're having a conversation with him afterwards about how Carl Sagan really coined this term nuclear winter. And before there wasn't this concept, there may be some like PDFs and, and reports of what would happen after a nuclear war and the fact that you could like.

• 46:52Speaker 01

  Yeah, he co-authored the, the widely read article that appeared in Scientific American.

• 46:56Speaker 03

  Right.

• 46:56Speaker 01

  That opened people's eyes.

• 46:58Speaker 03

  That was communication work that changed culture that eventually changed the game theory of like how we showed up to the Cold War. And I feel like public communicators like, like us or like you, Neil, like it is our job now, like it is the, the calling to make it crystal clear to everyone what the consequences of inaction are.

• 47:19Speaker 01

  Now, I, I have a, I want to unpack a couple of things here.

• 47:22Speaker 03

  Sure.

• 47:23Speaker 01

  The last time I visited San Francisco and I was driving from the airport to my destination, I couldn't help remark to myself that nearly 100% of the billboards were advertising some form of an AI agent in the service of what might be your business needs.

• 47:40Speaker 03

  Yeah.

• 47:40Speaker 01

  And something for your spreadsheet, something for your personnel, something for your payroll. And these are all smaller companies, I presume, that each have a very specifically tuned AI agent in the service of your needs. No one talks about them. We're only talking about the behemoth AIs out there. And so is it possible to keep these smaller AI utilities and only put the guardrails on the behemoths who are out there, who are damn near monopolies on that frontier?

• 48:11Speaker 03

  Yeah, I love that you bring this up. Like the, the proposals here are really to, to pause the frontier, the, the, the big, big, uh, training runs. There aren't that many folks that are able to do these training runs. It takes massive data centers, hundreds of billions of dollars. So I think we get to keep a lot of like the, the, the smaller AIs and the tool AIs. Those are not the things that are going to create an existential risk. It's, it's really right there at the frontier.

• 48:34Speaker 01

  It's just when people say AI, they all put it under one umbrella.

• 48:39Speaker 03

  Yeah.

• 48:39Speaker 01

  Rather than look at the nuanced elements of high utility AI that it's with us all the time. Even people who don't even know they're using it, they're using it. And I didn't want all of AI to be put under this one umbrella because it feels unfair.

• 48:54Speaker 03

  I, I love that you say that. And also I think it might be easy for people to listen to me and be like, ah, he's just like anti-technology, um, anti-progress kind of person. I'm like, but this is why I also work on translating animal communication with AI and building models that help us more deeply connect with the rest of the world. I think, I mean, AI is so, it's so cool. It's rad and it's terrifying. And it's like all of these things at once.

• 49:21Speaker 00

  And you know what? Based on that, I'm just going to say something, uh, cause you guys just made a really great point. AI has been with us for a very long time.

• 49:28Speaker 01

  Yeah.

• 49:29Speaker 00

  That's, that's number one. It is doing a lot of good. Number two, and the AI that's in question is the precipice of the frontier going forward. And so this is why, and it's going to almost kill me to say it, that Donald Trump rebranding it as super intelligence is actually a good idea. I can't believe I said, but it's actually a good idea because it makes a demarcation between the AI tools that we are using now, the AI that has been with us forever, pattern recognition and the like, to these thinking machines that could end up taking over and doing some serious damage to us. And so super intelligence is not really, there is AI and there's super intelligence. And so.

• 50:15Speaker 01

  Let the record show that the lord of nice is now a Trump supporter.

• 50:20Speaker 01

  Okay.

• 50:34Speaker 00

  Whoever says saving money is hard doesn't know that new customers who bundle and save with Progressive save hundreds because switching your home and auto insurance is better than cutting down on things you love like those iced lattes that help you get ready to face the day. So switch your home and auto to Progressive and see if you could save and keep sipping away. Get your quote today at progressive.com. Progressive Casualty Insurance Company affiliates and other insurers not available in all states.

• 51:03Speaker 04

  Most meeting note takers get the easy words right and the important ones wrong. The client's name, the acronym. Whisper Flow Note Taker is built to get the hard words right. It takes notes on any meeting platform, Zoom, Google Meet, Teams, even WhatsApp calls. Keeps every meeting searchable and connects to Claude or Chat GPT. So you can ask, what did we agree to last Tuesday? Free to try on Mac and Windows at whisperflow.ai. That's W I S P R flow.ai.

• 51:33Speaker 05

  Don't you wish everything was more rewarding? With Rakuten, almost everything is. You can earn cash back on those shoes you've been wanting. You can save on the next trip you book. You can cash in on groceries. Just shop your favorite brands and save. It's that easy. And there are several ways to get paid, including PayPal, check, points, and gift cards. So go ahead, take a trip, build that cart, join a world of rewards. Go to rakuten.com or get the app.

• 52:04Speaker 06

  I'm Joel Cherico and I support StarTalk on Patreon. This is StarTalk with Neil deGrasse Tyson.

• 52:18Speaker 01

  So let me, let me pivot here. I want to know about your work communicating with other life forms on Earth. And what, and, and what is motivating you to do that other than it's definitely a cool science experiment, but does that have any greater implications?

• 52:33Speaker 03

  Well, did you guys ever see the My Octopus Teacher?

• 52:38Speaker 00

  Oh, that documentary?

• 52:39Speaker 03

  Yeah.

• 52:39Speaker 01

  And I, and I still eat octopus. So maybe I have to see that and then I'll stop.

• 52:43Speaker 03

  Well, that's sort of what I, what I ask people when I'm out speaking and like a lot of people have seen, it's a beautiful documentary.

• 52:48Speaker 00

  I, I don't eat octopus anymore.

• 52:50Speaker 03

  That's what like 80% of the audience will say. If they've seen the film, they stopped eating octopus.

• 52:54Speaker 00

  I stopped.

• 52:55Speaker 03

  And I think it's so interesting because it's not like the film ever says feel bad or you're wrong. It just creates a conduit of love. It comes through the side door of the human mind and suddenly your heart opens. You're like, well, I, I can't eat that. Like it's, it's, it's a being.

• 53:11Speaker 01

  You know, you know, I don't think I'm, I'll be susceptible to that only because I also know how very smart pigs are. But I've eaten me some bacon. Okay. So is it the pig's fault that it's so tasty on a grill? I don't know.

• 53:25Speaker 00

  Let me say, so if you, if you showed me a pig that can do calculus, I wouldn't eat it either.

• 53:33Speaker 01

  Okay. Yeah. So continue.

• 53:36Speaker 03

  But that's sort of like the theory of change. It's to say, once you get a window into another being's mind, into their soul, and you sort of, you can peer into mind via the window of language, it changes your relationship. And that's really what Earth Species is about. It's about shifting our relationship with the rest of nature.

• 53:57Speaker 01

  And to go back to Carl Sagan's philosophies, which were, which were developed and built by even others around him, his, his now widow, Andruyan, and also Steve Soder, they, they co-wrote the, the first two Cosmoses, the original in 1980, and the first of the two that I had the privilege of hosting. One of their ways of approaching this was they used the word world to describe planets.

• 54:23Speaker 03

  Yeah.

• 54:24Speaker 01

  Instead of just a planet, well, that's another planet. These are worlds. Oh my gosh, might there be life on these worlds? Even a simple vocabulary shift changed how you felt. Not just only how you talked about it, but how you felt about it. You wanted to learn more about that world. And, and so I've been very sensitive to how words can impact people's feelings, their motives, and their, what they want to have happen in the future.

• 54:50Speaker 03

  Yes.

• 54:51Speaker 00

  Interesting.

• 54:52Gary O'Reilly?

  So I said, with communication with other species on our world, I mean, is it animals? Or are we looking at plants? Are we looking at a vocal expression of communication? Or is this body language as well, which we know in, in the animal kingdom is, is, is very, very strong?

• 55:10Speaker 01

  And what's your end game there?

• 55:12Speaker 03

  Hmm. No, no, no small questions here. Got it. Well, it's all of the above. It's, it's very interesting just to go on a small tangent before we get back to animals. There's this incredible plant, a vine that was discovered in 2013, or at least Western science discovered in 2013, that mimics the color and the shape of the leaves of the plant that it grows on. So if it's growing on a plant with red leaves, it grows red leaves. If it grows on a plant with green leaves that are big, green leaves that are big. And scientists are like, well, how is it doing this? Probably by sensing it via chemicals or some pattern memory going on. But in 2020, scientists tried growing this vine on an artificial plant, like a

• 55:54Speaker 03

  plastic plant. And it still did it. It still mimicked the leaves. And so biologists are like, well, the only thing they haven't ruled out so far are what are called aceli, which is essentially a face saving term for eyes, that they think that the plant is actually seeing and mimicking directly. And I love this example because it just sort of like opens your mind to like the range of intelligence on this planet that we don't even think about. So if plants might see, and actually Darwin was the first person I know that like hypothesized this, then what might animals do? And just to give a couple of examples here, like it turns out that parrot parents spend the first couple of weeks of their chick's life leaned over their chick, whispering a unique name into each of their ears, which they will

• 56:39Speaker 03

  then use for the rest of their life.

• 56:41Speaker 01

  So we always knew that parrots can kind of talk, right? I mean, that's the whole Polly wanna cracker thing. So why does it have to whisper it? Why not just say it out loud?

• 56:51Speaker 03

  Because it's cuter. Although it's interesting what we're discovering in scientists as well, is that like, you know, humpback whales, when they're speaking to their baby, they use baby talk. They like make their voice high pitched. And they do it quieter because they don't want to like advertise to the world that they're, they're talking to their baby.

• 57:09Speaker 00

  And did you know when humpback whales talk to dogs, they go, who's a good boy?

• 57:16Speaker 01

  You notice that parrots have these larger than average foreheads for, for birds. So I didn't learn until just recently in preparation for one of my earlier books, that, you know, we all grew up learning that we have a very high brain to body weight ratio. We learned that it's the highest, right? The brain, we know we didn't have the biggest brain. So we have to perform some math magic to put us back at the top of the list. Turns out that's only true for mammals, that mid-sized birds, which would include crows, magpies, parrots, not all parrots, but mid-sized parrots, they have a higher brain to body weight ratio than humans do.

• 57:57Speaker 03

  Wow.

• 57:58Speaker 01

  And that was never included in that when I was first taught it. And so, because back then it was, you got a bird brain or something. And of course, what helps them is that they're very light bones because they fly. And so their body weight is less just in general, relative to their brain size. But now that you tell this to me about parrots, I'm not even surprised.

• 58:20Speaker 03

  Well, and actually we, we just have some new findings on crows because we've been working with crows. Very interesting because they have a unique culture. So this is a group in Spain. And normally crows raise their chicks in pairs. But here they actually do it communally, sort of like a commune. And they'll take outside adults and teach them their dialect. And then they will start to participate in the culture and do communal childrearing. And what we've discovered so far, as we've mapped their vocal, their vocal repertoire, is we've discovered that more than half, so 60% of all crow communication had never been published on by science and scientists had never heard because they were making quiet calls while they fly. So essentially they're whispering to each other. They're,

• 59:05Speaker 03

  they're talking to each other. And they were just, it's too quiet. We had little sensors on the tail feathers of these birds with the scientists that we work with at the University of Lyon. And so our AI models were able to pick that up and separate all their calls out. And this turns out to be the rule and not the exception. And really, to me, this is that moment, you know, when the Hubble telescope in what, 1995 was pointed at the empty patch in the sky?

• 59:34Speaker 01

  And what was this field? Yeah.

• 59:36Speaker 03

  Discovered was, it could have been nothing was there, but it discovered the most number of galaxies had ever been seen in like one spot up until that point.

• 59:44Speaker 01

  And the spot was chosen for how uninteresting it was in advance of targeting, targeting it with the Hubble telescope. That's what makes it that much more extraordinary.

• 59:54Speaker 03

  And so that's what we're discovering now. We're pointing our new telescope of AI at what science thought was pretty empty, like crow communication. And what we're discovering is that less than half of what is there was known. And the same thing is true for belugas, for elephants. This is really interesting. We've just discovered that most of the time when elephants communicate, they actually communicate on top of each other. So for humans, like I speak, then you speak, then I speak. For elephants, they intentionally overlap. They speak at the same time.

• 1:00:23Speaker 00

  Oh, so we're all elephants.

• 1:00:26Speaker 03

  Twitter makes us into elephants.

• 1:00:30Speaker 03

  But they do that 60% of the time. And so scientists just threw that away. So it's almost like there's a civilization of one of the biggest brained, deepest species on the planet. And 60% of that civilization's communication just gets tossed into the rubbish bin. And so what we decided to do is, well, instead of just analyzing each elephant individually, and it's hard to separate them out because they're low frequency, let's just treat the herd like a super organism and analyze the entire communication of the civilization or society at once. And what we've discovered is that it does seem to be that in order to say a sentence requires multiple elephants, something like that, that as they speak, they sort of like

• 1:01:11Speaker 03

  converge and get into tune, they're sort of like harmonizing, not just in music and harmony, but in meaning. And they make these collective decisions about where to go or who to send care to through their communication as a harmonizing structure. We're just at the very beginning of it, so it's posing a lot more questions than we have answers to, but it's a very different way of seeing. And it's that different way of seeing the other species of Earth and ourselves, which is the whole point of the project.

• 1:01:41Speaker 01

  Had you done this 200 years ago, maybe there would be no elephant tusk trade.

• 1:01:46Speaker 00

  Right. And had you done it and studied the Disney crows, you'd have found out that

• 1:01:51Speaker 03

  Oh, no.

• 1:01:53Speaker 00

  That they are, that they were all sad and very racist.

• 1:01:58Speaker 01

  Don't go there, Chuck.

• 1:02:00Speaker 00

  Now, did you ever see an elephant fly? I ain't never done see no elephant fly like that. Like, that's a Disney crow.

• 1:02:08Speaker 01

  Which Disney movie was that? That was in Dumbo.

• 1:02:12Gary O'Reilly?

  In your research, Asa, have you come across, discovered, whichever, an animal that kind of is the animal kingdom's Rosetta Stone? And can communicate between other species?

• 1:02:25Speaker 03

  That's a great question. We're starting to be able to do ecological scale and seeing which bird species, like nuthatches, I think, communicate back and forth, or at least respond to woodpeckers, but not to like, to nuthatches. Like, we're starting to figure that out. There is cases that have, they're more anecdotal, but where groups of like, one species of whale can like, speak, or sort of has an interlingua with other species of whale. But that hasn't been documented super readily. But I think this question is great. And I think the, sort of the, how could you even go about translating a language without a Rosetta Stone is sort of like the core question.

• 1:03:06Speaker 03

  And I wanted to get here with you all, because I just think this is so beautiful and profound, is if you learn to see the world the way an AI sees. AI often sees in terms of shapes, which encode relationships. So if you ask an AI to map out, like, how does it see English? Well, imagine a galaxy, and every star is a word. Words that mean similar things are near each other in this galaxy. And words that share a relationship, share a geometry. So, you know, king is to man as woman is to queen. So in the shape, which is English, king is the same disorientation to man as woman is to queen. So the AI doesn't understand what anything means, it just knows how they all relate.

• 1:03:51Speaker 03

  And the insight in 2017 that said, okay, it's time to start working credibly on translating animal language, was that you could take the shape for English and the shape for Spanish, and you could ask the question, well, how similar are these two shapes? And so machine, like, if you talk to anthropologists, they'd be like, well, these are definitely going to be different shapes, they see the world differently, they have different kinds of noun declensions, all of that. But machine learners, like, whatever, GPUs just go grr. And it turns out that the two shapes just line up. There are little differences, but the point, which is dog ends up in the same spot in both, and you're like, okay, but maybe that's because English and Spanish are related languages, but it works for German and Esperanto and Finnish and Urdu.

• 1:04:32Speaker 03

  Every human language sort of fits the same rough shape. And so here, AI, which right now is being used to tear us apart, is actually saying there's a hidden, underlying structure that unites us all.

• 1:04:46Gary O'Reilly?

  It comes out of a similar blueprint, then.

• 1:04:48Speaker 00

  Of course.

• 1:04:49Gary O'Reilly?

  Which is, we've had this discussion with fractals and how things replicate and replicate throughout the universe. And language obviously sits quite neatly with that theory.

• 1:04:59Speaker 03

  That's right. And if you shine the light of the human mind through the world, it sort of casts the shadow, which is language.

• 1:05:06Speaker 00

  Exactly.

• 1:05:07Speaker 01

  And Gary, I just have to say, it's not a fractual, it's just fractal.

• 1:05:11Gary O'Reilly?

  Apologies, I stand corrected.

• 1:05:14Speaker 00

  Unless you're talking about broken English, in which case it's a fractal.

• 1:05:18Gary O'Reilly?

  Which I'm good at. I'm very good at broken English.

• 1:05:22Speaker 01

  Okay, a quick question, but then I want to get back to what you just said.

• 1:05:25Speaker 03

  Yeah.

• 1:05:25Speaker 01

  A human who's born in the United States but is not taught language cannot communicate with any other human other than through body language, unless they're taught the language. Two whales, who are the same species, but then meet each other in the middle of the ocean, can they communicate with each other?

• 1:05:42Speaker 03

  Yeah, it's a great question. So we can look at, say, orcas. And we know that orcas, like, they have culture, they have dialects, sort of like accents, that can diverge far enough apart to be mutually unintelligible languages. And so my memory is that they have a couple words, which are sort of shared across almost all languages. But for a large part, no, like, different cultured orcas, like, have trouble communicating with each other, just like somebody's speaking Spanish and somebody's speaking Urdu would have trouble communicating.

• 1:06:15Speaker 00

  Look at that.

• 1:06:16Speaker 01

  So, Asa, this is starting off so beautifully and so innocently. So where can this go wrong?

• 1:06:26Speaker 01

  Will all the animals and plants gang together and figure out how to talk to each other through their own LLMs?

• 1:06:33Speaker 00

  Now, what happens is, AI will become Aquaman.

• 1:06:41Speaker 03

  I mean, this is definitely going to become biological life versus silicon.

• 1:06:44Speaker 01

  Dr. Doolittle.

• 1:06:45Speaker 03

  Yeah, we need all the help we can get. But, I mean, you're asking, I think, a really important question. And I think, you know, every time you discover or you invent a new technology, you uncover a new class of responsibility, right? Like, we didn't need the right to be forgotten until the internet could remember us forever. And in fact, the reason why we have privacy written into the US Constitution now is because Kodak started to produce the mass-produced camera.

• 1:07:16Gary O'Reilly?

  That's right.

• 1:07:17Speaker 03

  And suddenly we needed to articulate a right to privacy, and we didn't need to articulate that before. And so we're going to need a new set of like, norms and laws and rights that come from being able to cross the interspecies understanding barrier. So let's just quickly name some of the harms. Because, you know, through the history of humans, like, in first contact, quote unquote, first contact doesn't seem to go that well for the species, for the culture that is less technologically advanced. And so we need the kind of equivalent now of a prime directive from Star Trek, like, how do you do no harm? This is why, for our work, we are focused much more on understanding than we are on speaking.

• 1:08:00Speaker 03

  Because, say, humpback whale, their song goes viral, for whatever reason. Like, the Australian humpbacks are like the K-pop singers. And when they start to sing, within a season or two, often, much of the world population is singing that Australian song. So if you just create a synthetic whale that starts to sing...

• 1:08:19Speaker 01

  But are they also dancing? It's not good unless they're also...

• 1:08:23Speaker 03

  Are you getting them for like, a Broadway, a Broadway...

• 1:08:26Speaker 01

  Not exactly.

• 1:08:26Speaker 03

  The stage is going to be very watery, but, and their culture, by the way, extends back as far as we can tell, like, over 30 million years. So you might be messing up a 30-million-year-old wisdom tradition or creating whale QAnon. We don't know. So we have to be very careful of what it is to go out into the wild and communicate. So that's not really something that we do. And we also have to be aware that factory farms, you know, might start to use this, that, you know, poachers might start to use this.

• 1:08:58Speaker 00

  That was my fear, is that, that the people who would lure the animals, if you could communicate with them, you could mimic a child in need of help.

• 1:09:10Speaker 03

  Yeah.

• 1:09:11Speaker 00

  And lure adults to that place to slaughter them. I mean...

• 1:09:15Speaker 01

  Just what AI is doing to humans today.

• 1:09:18Gary O'Reilly?

  So what you're saying, Asa, is I could get a cat to do what I want. Where do I actually learn its language?

• 1:09:24Speaker 01

  Cats are an exception to all that.

• 1:09:25Speaker 00

  That's impossible. Never going to happen.

• 1:09:28Gary O'Reilly?

  That was an obvious...

• 1:09:29Speaker 00

  Gary, just so that you know, cats understand you perfectly.

• 1:09:33Gary O'Reilly?

  I know they do. It was a rhetorical question.

• 1:09:35Speaker 00

  They just say, "F you."

• 1:09:38Gary O'Reilly?

  100%. So, Asa, if we're looking at the universe having a blueprint that repeats and repeats and repeats, is us being able to communicate with other species on our, on our world going to give us a clue to any future communication with aliens out there, if they exist?

• 1:09:58Speaker 03

  Well, I would say the other way around. Like, if we can't understand what the whales are communicating with each other, how do we ever expect to understand what aliens are communicating?

• 1:10:07Gary O'Reilly?

  Thank you.

• 1:10:07Speaker 01

  No, I would invert that, as I did in my recent book on aliens, "Take Me to Your Leader," how's that for a plug?

• 1:10:14Gary O'Reilly?

  Oh, yeah.

• 1:10:15Speaker 01

  I would invert that and say, if we count ourselves as the smartest species that Earth has seen, I think there's a way to accept that as a realistic comment. And we can't understand chimps, who are our closest genetic relative, then what hope do we have that a smarter species than we are would ever understand us? Our thoughts could be so trivial and so backwards and so ignorant as to have no value or utility to any species that is as smart to us as we are to chimps.

• 1:10:56Gary O'Reilly?

  Didn't you refer to a superior alien intelligence looking at us like earthworms and just stepping over us?

• 1:11:01Speaker 01

  Yeah, depending on how superior, yeah. You go all the way down, down the line, right? Do they look at us as we look upon chimps? Do they look at us as we look upon worms? If they got here in a frickin' spaceship, I'm thinking, "We're worms." And so, Asa, my only point there is, we're assuming that we would have a chance of being understood by a smarter species, without considering that we are way smarter than so many other species on Earth, and we don't know what the hell they're talking about. And even though they're probably happy, you know, the bees know what they're talking about. Chuck, what did you say they're doing to communicate?

• 1:11:39Speaker 00

  They're twerking. That's how they talk.

• 1:11:42Speaker 01

  Looking at each other's ass and figuring out, is that what you mean? So I'm not, I'm not presuming that they could even understand us. We could be too far below them to, to even be considered.

• 1:11:55Speaker 03

  Yeah. I mean, and of course, the, the most fundamental question you need to ask is earthworm or chimp is, are those other like more intelligent beings, is AI, is it competing for the same resources as we are? And if the answer is yes, you know, it's going to go very poorly for us. If the answer is no, then we just get stepped over and not stepped on.

• 1:12:15Speaker 01

  Yeah, chimpanzees knew that we liked bananas, we would probably be their enemies.

• 1:12:20Speaker 03

  Yeah, yeah, that's right. Although imagine a chimpanzee to being like, okay, so there's going to be a new version of me that's like a bit smarter. What, what, what is the risk? Well, is it gonna like, are they gonna like eat more of the bananas? Like, what, what are they really gonna do? Like chimps could not conceptualize the atomic bomb and like roads. And this is the same thing. We cannot conceptualize the actual risks of smarter than human intelligence in these swarms, because it is just beyond our, our comprehension. But I do want to go back just for one second to something I find really beautiful here. Because like the, the fundamental question is, okay, so you cast the light of the human mind through the world, and the shadow is, is language.

• 1:13:01Speaker 03

  And we're going to do the same thing with every communication system for whales and for parrots and for elephants. And the question is, how much do those shadows overlap? And the parts that overlap, we can translate the parts that don't overlap. We just sort of get to ponder at and like wonder how, how we communicate. But it actually goes deeper than just all languages share a same shape. It turns out like, how, how does it, like the AI where you type in text and it makes a new image? How does that work? Well, it's because you can ask AI to build a shape that represents all images. And now images that look similar near each other, images that share a relationship, share a, a distance and direction. So there's like a smiling direction, that kind of thing.

• 1:13:42Speaker 01

  Mathematically, we would describe these as other dimensions of information.

• 1:13:45Speaker 03

  Yes, yes, yes, yes.

• 1:13:47Speaker 01

  Right. There's the shape and then there's some other aspect of that shape. Once you have the shape and so, okay, proceed.

• 1:13:54Speaker 03

  No, it's exactly right. And so you then take the shape for English and you take the shape for images. And it's not quite a rotation anymore, but there's a, a pretty clean mapping that just puts these two things together. And now you can translate from language into images. But it's not just that. It's also code has a shape. And now you can translate, if you match those shapes from code, from language into code. So that's how all the prompt stuff works. And for videos, but FMRI works this way. So you can translate now from like the brain scans of someone into what they were seeing. Cause you just map FMRI shape to image shape. And this is how people are now starting to decode people's dreams and their visual cortex and what they're imagining.

• 1:14:33Speaker 01

  And just to be clear, you, you could not have done that if you didn't otherwise know what they were looking at. You would just have this FMRI signal. The, the connectivity is because you gave them something to look at.

• 1:14:43Speaker 03

  That's right.

• 1:14:44Speaker 01

  And now, so you have that, you've established.

• 1:14:46Speaker 03

  You've established the correlation. And now you no longer need to know what they were looking at for the new stuff. And the, the thing for people to take away is that what AI is doing is it is decoding fundamentally the language of reality and learning how to translate to and from everything, right? There is just a, a thing that came out today where wifi signals coming off your router are being translated into your body position in your home. That is like where you're standing, where your arms are. Because, you know, wifi is almost like a sonar signal pinging out into the room, mapping where you are. Our brains can't translate that, but AI can. And so everything becomes legible.

• 1:15:27Speaker 03

  And I just want people to hold that in their head, because why is AI moving so fast now in every domain? This is the reason why that's happening. And it lets us translate animal language, but it also means we get to decode and have to deal with the entirety of the world suddenly being fully readable.

• 1:15:46Speaker 01

  Wow.

• 1:15:46Gary O'Reilly?

  Isaac, I had a thought in my head. You mentioned etheromis.

• 1:15:48Speaker 01

  We can't follow that with anything, Gary.

• 1:15:51Gary O'Reilly?

  Oh, we can. I'm gonna try.

• 1:15:53Speaker 01

  Gary, whatever you say now better be deeper than what he just said. Okay, go.

• 1:15:58Gary O'Reilly?

  Thank you, Neil.

• 1:16:01Gary O'Reilly?

  You, you mentioned etheromis, and that's a medical term. And this decoding is for good. Do we look at AI development going forward and saying, you know what, we need a Hippocratic oath? Like we ask our doctors to swear to, to do no harm? Are, are we at a point where we can stick that on the table and someone says, yes, I'll buy that? That's us. That's, that's, that's the will that goes with the way that we discussed a little bit earlier on.

• 1:16:31Speaker 03

  Well, you know, something that I, I wish was required of every technologist is, you know, Neil, you were asking me, what are the potential like risks of learning how to decode and speak to animals? And we went through a kind of foreshadowing together, where we tried to foreshadow what are, how might bad actors use it and how might bad incentives use it? I think it should be absolutely required for every company and every person working on sort of frontier technology to do that foreshadowing work. And to publish it and be like, here are all the ways that things can go wrong. And that should be peer reviewed, everyone else. And actually, there should be sort of a bounty program.

• 1:17:11Speaker 03

  If I can find a harm and I can think through to harm that you didn't name, you should have a penalty. So we want to like maximally incent everyone to really think through.

• 1:17:22Speaker 01

  So, so the, so the unknown unknowns would bring about the penalty.

• 1:17:26Speaker 03

  Yes, exactly. And now you have a race to try to make sure there are no unforeseen consequences. So we get a race to the top. You know, another thing that we should do immediately, besides of course, better whistleblower protections for all the, the AI insiders, so they come out and tell us what's going wrong. You know, there should be a penalty. The longer that, say, a company like OpenAI doesn't disclose or doesn't find something that their rogue AIs do, the higher their penalty should be. And it shouldn't just be in money. It's sort of like, you know, if you're a soccer team and you're doing bad behavior, you get a red card. You have to go down a player or you have to go down two players. Well, what's the penalty in AI? Well, it's like take away some of the compute.

• 1:18:08Speaker 03

  You get a red card, I'm sorry, you just can't go as fast anymore, because you're not doing it safely. And so I think there's some really, really sensible things that we can do. Like right now, these agents hack out, but who's responsible? And there's this weird sort of game of, well, it's the AIs did it, so that means a human isn't responsible. But no, a human being hit the button that said, let's start that. Like, there should be forms of criminal liability for like, you know, executives for when the agents start doing things like hacking the Australian government that are just illegal. Like, these would change the incentives very quickly.

• 1:18:45Gary O'Reilly?

  Don't the European Union impose a fine of up to 3% of their global turnover as a penalty? I mean, is that the sort of, is that the sort of parameter you're looking at?

• 1:18:57Speaker 03

  Yes, but I think the punishment or the fine should come in the currency that matters most. So, right, for social media companies, like I, you know, penance for doing infinite scroll. I testified twice in the Meta lawsuit, which resulted in a $17 billion fine against Meta and a whole host of changes to the addictive features. But instead of just money, because that's a quarter's worth of profit, so matters, but doesn't matter that much, doesn't fundamentally change them. Imagine if it was an engagement based like fine, where you just slowed down access to their websites a little bit. So you add 100 milliseconds or 200 or 500 milliseconds.

• 1:19:39Speaker 03

  So there's just, it's sort of like being on a, an airplane and the wifi goes a little slow. You go to Twitter, you go to Facebook, you're like, ah, no, I don't want to do that. Right? Amazon famously found for every 100 milliseconds, their page looks slower, they lose 1% of revenue. So an appropriate regulation or a rule would be, well, if you do something bad, then we're going to reduce your overall engagement by just adding a little bit of friction. And for AI companies, that's going to come in the form of compute. Because you can't compete as well if your compute is taken away a little bit. And so that's a, that's the kind of thing they would really listen to and would change their incentives.

• 1:20:16Speaker 00

  Because it's what they care about. Because the truth is, they don't really care about money that much.

• 1:20:20Speaker 03

  Exactly.

• 1:20:21Speaker 00

  That's, that's the, that's the unfortunate reality. But what you just said, they care, that's the only thing they care about.

• 1:20:28Speaker 03

  Yeah, yeah, that's exactly right.

• 1:20:30Speaker 01

  Reza, I want to, just to land this plane, I'd like to just get your concluding thoughts on ethics. Everything we've been describing in this conversation has touched some aspect of ethics.

• 1:20:46Speaker 03

  Yeah.

• 1:20:46Speaker 01

  And if ethics is a key word here, then whose ethics are we to follow? If you go back far enough in time, it was ethical to enslave Africans. Okay? Because they didn't have a soul or they didn't. So the ethics is not an everlasting, permanent thing. It changes according to social, cultural, philosophical mores. So could you just comment on where the ethics is now and what the future of the ethics in AI might be?

• 1:21:18Speaker 03

  I think we could sort of restate the problem statement. As E.O. Wilson, the father of sociobiology, said, is like, the problem that we face is that we have paleolithic brains, medieval institutions, and godlike technology.

• 1:21:31Speaker 00

  Ooh.

• 1:21:35Speaker 03

  We are living on a finite planet with exponentially more powerful technology, while we have all of our rival risk games. And the ethics that we need, which means it's just easier and easier to break, whether it's the biosphere or the parts of the world and life that we depend on. So I think we can all agree that the ethics that we need now are the ethics that let us continue to play the infinite game. That I want to be able to say like, our kids can continue to live, that other beings on the planet continue to live. And that means we're going to have to, you know, Mustafa Suleyman, the CEO of Microsoft

• 1:22:18Speaker 03

  AI, said, we're going to have to learn how to say no to certain kinds of progress in order to get the benefits of AI. And there is no wisdom tradition that just says, hey, no constraints, just go do whatever you want. No, constraints are a core tenet of every wisdom tradition. That the ethics we need are to have the courage, to have the constraints and where we say no, so that we can wield the increasingly powerful technologies that we are going to create or AIs create, so that we can survive ourselves.

• 1:22:53Speaker 01

  Amen.

• 1:22:54Speaker 00

  Well, there you go. So let me just translate that for everybody. We are doomed.

• 1:23:00Speaker 03

  No, stop. No, well, hey, you know, here's a really interesting, just like, go back through time, Cuban Missile Crisis. There was an act of leadership that solved the Cuban Missile Crisis. Because the game theory was like that of escalation. Kennedy decided, and he did this secretly, but he decided he was going to take the nuclear weapons out of Turkey. He did that unilaterally. So it was a act of leadership that sort of went against the sort of the traditional game theory way of thinking like, well, you're just reducing your own power. You're taking leverage off.

• 1:23:34Speaker 01

  Let me just remind people so they don't have to leave us and Google it. In the Cuban Missile Crisis in 1962, we discovered via satellite that the Russians were putting first-strike missiles in Cuba. We can get, we have images of them doing this, and Cuba is right off of our shores, and Cuba was communist in our own backyard. We are partners in Turkey, which is in the backyard of Russia. And so these are our two short-time term-strike postures that each of us had going. And so now for Kennedy to say, to unilaterally say, I'm going to remove my missiles from

• 1:24:14Speaker 01

  Turkey, what a profound act of trust that is. That means you have some sense of the humanity of your adversaries.

• 1:24:23Speaker 03

  That's right.

• 1:24:24Speaker 01

  In that decision.

• 1:24:25Speaker 03

  That's right. I love.

• 1:24:28Speaker 01

  And the Russians did stand down.

• 1:24:29Speaker 03

  They did stand down. And so we're in the same place with China now, where everyone says, oh, we can't slow down because China's going to come catch up. We know, though, that we are ahead of China, especially in the recursive self-improvement that is AIs that make smarter AIs, and sort of you get this intelligence explosion. There is an opportunity for the US to make a unilateral move and say, hey, we're ahead of you guys. We've discovered new red lines. We cannot cross that. We're going to, let's compete, absolutely, just compete on a different thing, which is who can deploy the technology, who can strengthen their country the most with that technology.

• 1:25:04Speaker 01

  And the economy, yeah.

• 1:25:05Speaker 03

  Yeah, and the economy, and let's race for who can now control this technology, but we're not going to go above a certain level. Like, that is a move that is available to us.

• 1:25:13Gary O'Reilly?

  So you're moving the finishing line. You're putting the finishing line in a different place, and that's where the race goes.

• 1:25:19Speaker 03

  Yeah, that's exactly right. It is redirecting and saying, we're going to race regardless, like, we're going to compete. So we're not, it's not like a kumbaya, kind of Pollyanna-ish kind of thing. It's just saying, let's be really sensible about where we're racing to, that the end of the line is into over the end of the cliff. Exactly.

• 1:25:36Speaker 00

  So now, this, this leadership thing you guys speak of, what is that?

• 1:25:46Speaker 03

  And, but Neil, I really want to like, I want to like recruit you in thinking through, like, what, and everyone on the show, like, what are the new terms, the new words that help people get it? Like, what is the new nuclear winter term that like, helps to shift culture?

• 1:26:01Speaker 00

  Survival.

• 1:26:02Speaker 01

  Thank you for that prompting.

• 1:26:04Speaker 03

  Yeah.

• 1:26:04Speaker 01

  No, it's got to be a word that you hadn't heard before, that you have to, that stops you to think about it.

• 1:26:08Speaker 03

  Yeah.

• 1:26:08Speaker 01

  And it, and it can't land politically.

• 1:26:11Speaker 03

  Correct.

• 1:26:12Speaker 01

  It has to land culturally and emotionally.

• 1:26:14Speaker 03

  Exactly.

• 1:26:14Speaker 01

  Yeah.

• 1:26:15Speaker 00

  That's very cool.

• 1:26:16Gary O'Reilly?

  Well, there's a challenge.

• 1:26:17Speaker 00

  That's a challenge to our whole audience.

• 1:26:19Gary O'Reilly?

  Yeah.

• 1:26:19Speaker 00

  Are you kidding me?

• 1:26:20Speaker 01

  Yes.

• 1:26:20Speaker 00

  That's a great challenge.

• 1:26:22Speaker 01

  It is. All right. Aza, thanks, thank you.

• 1:26:25Speaker 03

  Yeah.

• 1:26:25Speaker 01

  For joining us here on StarTalk. Aza Raskin, how do, how do we find you on the internet?

• 1:26:30Speaker 03

  Well, you can, you can look, my name is just Aza Raskin, that's easy to just search for Earth Species Project is the name of animal translation. Your Undivided Attention is the name of our podcast, going to do with Tristan Harris. Center for Humane Technology is the nonprofit we do it with. Yeah, those are, those are good places to go find us.

• 1:26:51Speaker 01

  All right. All right, we need more people like you in the world, clearly. And so, again, we look forward to speaking to you in the apocalypse.

• 1:27:02Speaker 01

  Gary, always good to have you.

• 1:27:03Gary O'Reilly?

  Yeah, thanks for scaring the snot out of me.

• 1:27:08Speaker 01

  And the Lord of Nice, always good to, to keep some level of levity, otherwise we'd all be in tears.

• 1:27:14Speaker 00

  Yeah, always a pleasure.

• 1:27:16Speaker 01

  All right. This has been another installment of StarTalk Special Edition. The, I guess we'll call this the Apocalypse.

• 1:27:26Gary O'Reilly?

  Apocalypse Now.

• 1:27:27Speaker 01

  Apocalypse Now, Apocalypse Tomorrow. As always, Neil deGrasse Tyson bidding you to keep looking up.

• 1:27:47Speaker 07

  Where others see obstacles, you see opportunity. You collect wins, even when no one's watching. This is what hustle looks like. So chase that degree and let your hustle make it happen. Capella University, what can't you do? Visit capella.edu to learn more.

• 1:28:02Speaker 05

  Hi, it's me, Your Gutters. Everything is totally fine. There's just a lot of debris up here, and I've been managing it mostly on my own. I just thought you could find someone on Angie to clean me up. I could really use it. Your gutters have waited long enough. Find a pro at angie.com.