Artificial Intelligence Ethicology (WILL A.I. CRASH OUT?) with Abeba Birhane

Who’s babysitting AI? Will it steal your job? What happens when you’re rude to a chatbot? Cognitive scientist, Trinity College professor and Artificial Intelligence Ethicologist Dr. Abeba Birhane lets me ask her not-smart questions about legislation around AI, auditing datasets, environmental impacts, booby traps, doorbell narcs, commonly used fallacies, how the “godfathers’ of AI feel about their creation, robots doing your homework, and and whether or not AI is actually the root of all evil. Also: bacon ice cream and why Siri is a girl.

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

• 0:00Alie Ward?

  Peppa Pig is your preschooler's best friend. Give the gift of bike riding confidence this year and help Peppa learn to ride her bike. Kids can imagine learning to ride a bike with the Ready Pedal Go Peppa Bicycle Playset. Put Peppa's helmet on, attach the push stick to her bike, and push to make her pedal. Ready Pedal Go Peppa and more from Peppa Pig are now available at all major retailers this holiday season.

• 0:29Alie Ward?

  Oh, trips are full of unexpected discoveries. Maybe you see a sea slug, or you find out that your friend still sleeps with a stuffed animal. But when it comes to your vacation rental, there should not be any surprises. When you book a Verbo vacation rental, if what you expect isn't what you get, Verbo Care has you covered. If something's broken or just not right, Verbo will work to make it right. Plus, you get 24/7 live human support with every booking. Grab your stuffy, we're going on vacay. So book today on the Verbo app. If you know, you're Verbo. Terms apply. See verbo.com/trust for details. This podcast is brought to you by Aura. Your body has a lot going on. Sleep, stress, recovery. You would be surprised how connected it all is.

• 1:13Alie Ward?

  Aura Ring tracks more than 50 metrics from your finger, and the Aura app brings that information together to help you see patterns across your sleep and stress, activity, and more, so you can start connecting the dots and figure out what works for you. Learn more at aura ring.com. Wear your best life with Aura. Oh, hey, it's the tag that you wish you'd cut out of your shirt, Ali Ward. And for every one of us that has seen AI become more and more present in our lives and wondered, is anyone driving this bus? I have here for you a chat with an expert who tells us exactly who is driving the bus and where it could be headed. Is AI evil? Does AI even care about us? Is it gonna kill us? Should we feel bad for it?

• 1:59Alie Ward?

  Don't ask me, I'm not a theologian. We're gonna get to it. Now, this expert is a senior fellow in trustworthy AI and an assistant professor at the School of Computer Science and Statistics at Trinity College in Dublin, Ireland. And they're cognitive scientists. They research ethics in artificial intelligence, and they've published papers with such titles as "The Forgotten Margins of AI Ethics: Toward Decolonizing Computational Sciences, The Unseen Black Faces of AI Algorithms, and the Values Encoded in Machine Learning Research." So, they're on it. And I got to sit down and visit and chat in person when I was in Ireland just last month. Also, just a pleasing aesthetic side note, they were born in Ethiopia but live in Ireland,

• 2:44Alie Ward?

  and this expert has the most melodic cadence, just like Bjork. I was mesmerized. We're gonna get to all that in a sec, but first, if you ever need shorter, kid-friendly episodes with no adult language, we have Smologies. They're episodes in their own feed, wherever you get podcasts. Also linked in the show notes, that's Smologies. Also, thank you to patrons for supporting Ologies and sending in questions ahead of time. You can join for as little as a dollar a month at patreon.com/ologies. Thanks also to everyone who's ever left a review for the show. They help so much, and I read all of them, weirdly, including this recent one by Remaly, who wrote, "This podcast will expand your horizons and help you indulge in your hyperfixations, both known and yet to be discovered."

• 3:29Alie Ward?

  Remaly also says, "Sorry it has taken me over five years to finally write a review." Remaly, anytime's a good time. Thanks for that. Okay, let's get right into artificial intelligence etymology. It's the ethics of machine cognition. Is it cognition? We'll talk about it. What does ChatGPT stand for? Why is Siri a lady? Can you ask a robot for a cheeseburger yet? What happens when you're rude to a chatbot? Also, how do artists prevent getting ripped off? How much energy does AI take up? When we all lose our jobs, booby traps, doorbell marks, commonly used fallacies, how the creators of AI feel about AI? What is hype and what is horror? What are the benefits of AI? What happens if you assign a chatbot your homework?

• 4:14Alie Ward?

  And whether or not AI is the root of all evil or a pocket pal? With embodied cognitive scientist, professor, scholar, and artificial intelligence ethicologist, Dr. Ababa Birhane.

• 4:46Speaker 01

  I'm asking questions that would take hours to unravel, and they expect you to answer it in like one minute, two-minute marks.

• 4:54Alie Ward?

  And they're rushing you to get off? So no worries.

• 4:57Speaker 01

  I am Ababa Birhane. She hers.

• 5:00Alie Ward?

  Great. And AI, you have been an expert in this field for a while, but I haven't known about AI for that long. How long have you been studying it?

• 5:11Speaker 01

  Technically speaking, I am a cognitive scientist. So I finished my PhD about three years ago in cognitive science. So halfway through my PhD, then around the end of my second year, I left the cognitive science department and joined a lab where people do a lot of testing, evaluating and testing of chatbots and various AI models.

• 5:36Alie Ward?

  And what is exactly cognitive science?

• 5:39Speaker 01

  Cognitive science is very broad. So traditional cognitive science tends to be about understanding cognition, understanding human behavior, understanding human interaction and so on. And cognitive science often is not taught at graduate level. It's either at a master's level or a PhD level because cognitive science is really interdisciplinary.

• 6:06Alie Ward?

  And Dr. Birhane says that cognitive science is actually a mishmash of disciplines, or what's sometimes called the cognitive hexagon, with sides representing philosophy, psychology, linguistics, neuroscience, artificial intelligence, and anthropology, according to some institutions. And anthropology can also mean social sciences. Different institutions will phrase it their own way, but broadly speaking, cognitive science is a lot of stuff.

• 6:33Speaker 01

  So the idea is you will take, you know, important or helpful aspects from these various disciplines. And cognitive science then allows you to synthesize, to combine these various theories, even computational models, to understand human cognition. That's the idea. So from philosophy, for example, you will learn how to question assumptions. You will go down into the various questions around what's cognition? What's intelligence? You know, what's human emotion? So philosophy really lends you the analytic tools. Same from neuroscience. The idea is you get to learn how the brain works and use it in a way to synthesize from all these different disciplines.

• 7:18Speaker 01

  So that's traditional cognitive science.

• 7:20Alie Ward?

  You can call cognitive science COGSI if you want to. And she says that she's in a really niche specialty within COGSI, and it's called embodied cognitive science, which isn't just about understanding the human brain.

• 7:35Speaker 01

  Whereas embodied cognitive science is moving away from this idea of treating cognition in isolation. Your cognition doesn't end at your brain and your sense of self doesn't end at the skin, but rather it's extended into the tools you use. Anything you do, you do it as an embodied self, as an embodied person. So your body, your social circle, your history, your culture, even your gender and sexuality. All these are important factors that play into, you know, your understanding, your cognition, your intelligence, your emotion, and so on.

• 8:13Alie Ward?

  And when you're studying cognitive science, how do you not use your brain to think about your brain all the time? How do you get out of your brain? You do it.

• 8:24Speaker 01

  Yeah, yeah. Yeah, I mean, you have to.

• 8:30Speaker 01

  You have to use your brain. I'm trying. So you are familiar, you know, with Descartes' famous quote, cogito ergo sum, so I think therefore I am. The idea is you can know who you are, you can know you are a thinking being, you can know that you are someone. So that is like using your brain to understand your brain, so to speak. Whereas, again, the emphasis with embodied cognitive science is that, you know, all that is really very individualistic. Even to confirm our existence, it's through conversations with others.

• 9:11Speaker 01

  So that's why embodied cognitive science emphasizes others and communities and your body as really important factors in understanding cognition.

• 9:20Alie Ward?

  So embodied cognitive science kind of goes downstairs a bit, and it considers how a person's body experiences the world and how the environment shapes thinking and perception. And cognitive scientists have this thought experiment that tickled me, and it envisions a little person in your brain interpreting inputs. But then who is in the brain inside that little person's brain? Like, does it have its own brain inside the little person's? It's like Russian nesting dolls. And there are just infinity little humans inside humans' brains to interpret what the brain inside the brain is braining. And this is called, endearingly, the homunculus argument.

• 10:02Alie Ward?

  And yes, embodied cognitive science is like it's more than a tiny person in your skull. And how can we truly understand artificial intelligence if we don't first grasp intelligence intelligently? And when we're thinking about who we interface with, with AI, Way back in the day, there used to be a search engine called Ask Jeeves. And it was like a butler who would find you the answers to things. And now we have Siri and Alexa. Why do you think with AI, they've gendered and they've personified these voices that are actually a huge network of artificial intelligence?

• 10:45Alie Ward?

  Is that to help our brains understand?

• 10:48Speaker 01

  It's a little bit of like a marketing strategy, but it's also a little bit of appealing to human nature. We tend to kind of gender and personify objects. So if you are interacting with ChatGPT, for example, we tend to just naturally, you know, treat it as another person, another being, another entity. On the one hand, as you said, these chatbots, you know, there is no intention, there is no understanding, there is no soul in these machines. They are just pure machines. But also the developers and vendors of these systems, they tend to market them as kind of personified entities.

• 11:32Speaker 01

  Because it's much more appealing to think that you are interacting with another sentient thing.

• 11:38Alie Ward?

  I always wonder if they made some of them female voices because we're more accepting, we're less threatened by females. We're socialized to have a mommy figure come and help us with something. It's not as threatening that it'll turn on us. And also, it's like, oh, a lady will get it for you, you know?

• 11:59Speaker 01

  Yeah, yeah, I mean, we naturally tend to think women are much more like nurturing and they have the role of helping you. So it is kind of related to the social norms that really dictate society. So it's really also to some extent leaning on that stereotype that if it's a woman, you know, it's approachable, it's there to help you.

• 12:25Alie Ward?

  And yes, I am far from the first person to notice that digital assistants tend to be ladies. And some historians and media scholars think that it starts early by hearing a female voice while you're cooking in the womb. And that's why people love a lady voice. Or that the first telephone operator in the late 1800s happened to be a woman with a great voice, and then it just stuck. But as these AI avatars start to have faces and personalities and pastimes and Instagrams, why do we see mostly younger, hotter avatars? Well, there was a 2024 article published by Reuters Institute, and it reached out to this multimedia professor, April Newton,

• 13:08Alie Ward?

  who noted that a gentle, well-modulated woman's voice is usually the default for AI assistants and avatars because, quote, "We order those devices to do things for us, and we are very comfortable ordering women to do stuff for us." And also, just a side note for me. Did you know that the word "robot", it comes from a Slavic root for forced labor or slave? It's creepy. And historically, humans have loved capable servants, but not too capable, or else you're just begging for an uprising. So the future, it's also the past. Am I rooting for the robots now? I don't know. When it comes to Siri or virtual assistants, how is that different than the AI

• 13:51Alie Ward?

  that's been ramped up in the last couple of years? Is Siri and Alexa, are those AIs? Or are those just search engines? Is Google an AI? We call some things AI and then other things just computing. What's the difference?

• 14:05Speaker 01

  Yeah, so what's really difference between, say, traditional AI, whether it's implemented as a chatbot or as a predictive system, generative AI that has really exploded over the past two years? Think of ChatGPT, Gemini and Cloud and so on. The fundamental difference is that, do I go into the technical details or remain clear of technical?

• 14:32Alie Ward?

  Yeah, no, you can give us some technical details.

• 14:34Speaker 01

  I'll try to keep it light. There is no way of explaining the difference without getting into reinforcement learning. A typical classification system is an algorithm that you give it massive amounts of data. Think of like a face identification system. And these days, data sets have to be really big. You might even have trillions of tokens of images of faces, and you train your algorithm like. This is a face, this is a face, this is not a face, this is a face, this is not a face. It's called machine learning. If you have succeeded in your training, then your AI should be able to tell you it's a face or it's not a face. So this is like a typical, you know, classification system.

• 15:17Speaker 01

  What we consider AI is a very broad term that encompasses so many different subcategories.

• 15:24Alie Ward?

  Just side note, how alone am I in not knowing that Apple virtual assistant Siri stands for something? Speech, interpretation and recognition interface? Did you know Siri was an acronym? I didn't. Also, Apple is reportedly freaking out behind the scenes about ChatGPT for the last few years, being generative and having chatbot capabilities and longer conversations in Siri. So apparently, a lot of their efforts in self-driving cars got shuttled over to their AI division. And a lot of people at Apple are like, I can't even speak publicly about this.

• 16:00Speaker 01

  The other big subcategory under the broad umbrella of AI is NLP or natural language processing. So this is the area that deals with human language. So you have audio data that you feed into the AI system. And the idea there is that, you know, you are building an AI system that even make predictions about human language. NLP tools would learn, for example, predictive text. So what the algorithm is doing is kind of predicting what the next word is likely to be. So it's just predicting the next token, the next token. So that's kind of traditional classification or predictive systems.

• 16:42Alie Ward?

  And that was machine learning or NLP, which is natural language processing. And those deal with visual data turned into tokens and they predict language, like when your phone knows you better than you know yourself. And it's heartwarming and it's scary. It's like love. Now, ChatGPT, for example, I didn't know this, but the GPT stands for Generative Pre-trained Transformer. And a transformer is this type of deep learning system and it converts info into tokens and can handle more complex processing from language to vision to games and audio generation. So it's definitely a step up from just simple prediction.

• 17:23Speaker 01

  Whereas over the past year with generative AI, these are AI systems that do more than classification, more than prediction, more than aggregation, they are called generative AI systems. They produce, you know, something new. So image generators, for example, you can put a text description as a prompt and the AI system will produce an image that resembles based on your description. The same with language systems. So ChatGPT, for example, you put in a prompt and it's able to produce new answers. So this is where the new, this is what's new with generative systems.

• 18:03Alie Ward?

  And the systems also need to learn which of the words in a language model are the important ones, which is part of the self-attention mechanism. And then they generate based on statistics, like what is the most probable way based on the data sets it's learned from, from completing a certain sentence or prompt?

• 18:20Speaker 01

  So the training data really has a really significant impact in what outputs, whether it's image or text, what outputs these systems can generate.

• 18:33Alie Ward?

  Let's get to the juicy part. So with any tokens, would that data be in tokens? Like you were saying, facial recognition might have a trillion tokens. Does AI kind of scrub what we know of impressionist art and science fiction and anime? Does it scrub it and grab a bunch of tokens so then it can have those as reference points?

• 18:55Speaker 01

  So the training data sets is one of the most contested issues in AI because data is constantly harvested. So like your search history feeds into some kind of AI one way or another. I don't know if you are like me. If you signed up for some kind of bonus point, you go to a grocery shop and you tap that. So that is kind of like in the background. That's kind of collecting your behavioral data.

• 19:24Alie Ward?

  So that data may breeze through infrastructure like Google and then just be on its merry way to a third-party aggregator or a broker or an AI company itself. And they just yum, yum, yum gobble up the data.

• 19:38Speaker 01

  This kind of practice puts a lot of data gathering for the purpose of training AI systems, either illegal or borderline illegal. Yes, because, you know, first of all, there is no consent. People are not even aware that, you know, their data is being used to train AI.

• 19:56Alie Ward?

  So that's just number one on the general level. Let's get to art.

• 20:00Speaker 01

  But you will also have noticed over the past year or two, the creative community, writers and artists are realizing that their work, their novels, their writing, their, their arts are being used to train AI systems, again, without any compensation or with very little compensation after the fact if they go and contest, you know, the use of their data. And because large companies that are kind of developing AI, you know, think of Google, DeepMind, Meta, OpenAI, Anthropic, they are businesses. They, they operate under the business model.

• 20:45Speaker 01

  They are, you know, commercial entities. Their objective is to maximize profit. My work is auditing, for example, large-scale training data sets. People don't have access to what kind of data set these companies hold. So we don't have any mechanism to kind of take a look, to scrutinize what's, what, you know, what's in the data set? Where does it come from? How large is it? These are the kind of questions that simply, there is just no mechanism for tech corporations to be transparent, to, to open up for us to have an objective understanding.

• 21:22Alie Ward?

  So big companies are like, no, you can't see it. But auditors, like Dr. Bhani and colleagues, can observe open source, publicly available, similar data sets as proxies or substitutes to kind of figure out what might be going on organizationally behind the locked Willy Wonka factory gates of the other big tech companies.

• 21:42Speaker 01

  So we do get an idea, but through proxy data sets. So to answer your question, I mean, if you are an artist, a writer, if you have produced novels and so on, it's very, very likely that your work is being used to train AI, but there is very little legal mechanism to actually have a clear idea.

• 22:01Alie Ward?

  Is it like they're stealing a bunch of ingredients and then making something, but they're like, you can't see our recipe? And you're like, you stole the ingredients? And they're like, you can't see our resume? We made something with it. That's, is that sort of how it's going?

• 22:15Speaker 01

  Yeah, so the, the fact that they are protected by proprietary rights as a commercial entity means that there is virtually no mechanism to, you know, to force these companies to open up their data sets. Of course, you can encourage them. You can kind of, you know, appeal to their good sides and so on.

• 22:37Alie Ward?

  Good luck.

• 22:39Speaker 01

  But yeah, there is no, there is no legal mechanism. There is no law or regulation that says you have to open access or you have to, you know, share your data set.

• 22:51Alie Ward?

  And why not, you ask?

• 22:53Speaker 01

  Of course, if that is to happen, then, you know, all these AI companies would go out of business. As it is, a lot of them are under a lot of lawsuits.

• 23:05Alie Ward?

  Oh.

• 23:05Speaker 01

  Yeah, you know, from, from Meta to OpenAI. OpenAI itself is under a lot of lawsuits, including from the New York Times. So the minute they open up their data, it becomes clear that a lot of people's suspicions, especially, you know, the, the creative and artist communities, the minute the data sets are open, why they are going to court?

• 23:27Alie Ward?

  Is there any recourse that artists or writers have? Like, is there anything they can do other than trying to open a really big expensive lawsuit?

• 23:36Speaker 01

  Yeah, so writers and creatives are organizing for class action suits. I know there are a bunch of class action lawsuits, both in the US and in the UK.

• 23:47Alie Ward?

  Now, last August, there was this landmark case, and it decided in favor of artists. It found that generative AI systems like Midjourney and DeviantArt's Dream Up and Stability AI's Stable Diffusion, that those were violating copyright law by using data sets of billions of artistic examples scraped from the web. And Getty Images sued Stable Diffusion for copyright infringement. And the evidence was like almost funny if it were such a bummer. But apparently the generative AI so relied on Getty Images that it started adding a blurry gray box to some of its AI output, which was learned from the iconic Getty Images watermark. Just embarrassing.

• 24:29Alie Ward?

  Now, in March of this year, a judge ordered that this lawsuit between the New York Times and OpenAI can proceed despite OpenAI begging it not to. And the newspaper, New York Times, along with a few other journalism outlets alleges that OpenAI scraped a lot of their work to train ChatGPT. So there are lawsuits, but there are also just huge glaring holes.

• 24:53Speaker 01

  But also in the UK, for example, they are considering a regulation that leaves massive loophole for intellectual copyright issues that leaves artists and writers with absolutely no protection at all. So people are really organizing on the ground and they have massive amounts of petitions signed and so on. But also there are technical, I don't want to say solutions, technical kind of remedies. So you can use data poisoning tools, for example. You have.

• 25:29Alie Ward?

  Oh.

• 25:30Speaker 01

  Yeah, so you have.

• 25:31Alie Ward?

  Like a booby trap?

• 25:32Speaker 01

  Like one of them is called Nightshade. So for images, for example, you would insert various adversarial attacks that will make the data unusable for machines because it's maybe a tiny pixel that's been altered. So that's not visible to the human eye, but it kind of messes with the automated system or how machines kind of use this dataset. So there are various tools like that.

• 26:00Alie Ward?

  Another type of AI booby trap is called a tar pit and it sends AI crawlers in this infinite loop where they just get stuck and they can't escape. And I just love to think of like the AI system on the toilet, just scrolling, just not being able to exit and get back to work.

• 26:18Speaker 01

  Even if you find technical solutions now, the big companies are likely to, you know, come back with another solution that makes your own solution defunct. So you, yeah, you really have to be adapting to that constantly. So I think a viable solution has to come from the regulation, from the legal space.

• 26:39Alie Ward?

  And how do you feel about, is it Sam? Oh, my gosh. Sam Altman. Thank you. I almost said Sam Edelman, which is a shoemaker.

• 26:46Speaker 02

  What is wrong with me?

• 26:49Alie Ward?

  How do you feel about kind of some recent changes? Like last May, I believe, he went before Congress in the US saying like, hey,

• 26:59Alie Ward?

  we better watch out. And now I think he was like at the inauguration. So we chatted about this in 2023's Neurotechnology episode because just a few months prior, in May 2023, Sam Altman was in the news a lot as a cautionary voice. And as we said in that episode, if you're wondering why this was a big deal, Sam Altman is the head of OpenAI, which invented ChatGPT. And in spring of 2023, he spoke at the Senate Judiciary Committee Subcommittee on Privacy, Technology and the Law Hearing called Oversight of AI: Rules for Artificial Intelligence. He also signed a statement about trying to mitigate the risk of extinction. And he told the committee that, quote, "AI could cause significant harm to the world."

• 27:44Speaker 03

  My worst fears are that we cause significant, we, the field, the technology, the industry, cause significant harm to the world. I think that could happen in a lot of different ways. I think if this technology goes wrong, it can go quite wrong. And we want to be vocal about that. We want to work with the government to prevent that from happening.

• 28:03Alie Ward?

  And ultimately, Altman urged the committee to help establish a new framework for this new technology. And though in 2016, Altman declared that Donald Trump was terrible, he recently backpedaled on that. And Altman said that he's changed his mind and donated $1 million to Trump's reelection campaign in 2024. So Altman's thoughts on AI regulations likely have pivoted in the last few years since that hearing. It doesn't seem like regulations are going to happen very fast.

• 28:35Speaker 01

  Yeah, so this idea of, oh, we gotta watch out because our AIs might become sentient and might be out of control, might cause existential risk and lead to human extinction. So unfortunately, this is a very popular and commonly disseminated worry emerging out of AI. But there is very little scientific evidence for it. People have done a thorough analysis how such a possibility is just zero percent likely. There it's just, yeah. But human beings can be so terrible and they're learning from us.

• 29:19Speaker 01

  Yeah, so there is no intention, there is no wish, or there is no desire to act on something, to do something. But at the end of the day, it really is a massive, complex algorithm, of course. And that is to, to some extent unpredictable. But that doesn't mean, you know, AI systems, as they develop further, you know, all of a sudden develop intentionality, or wishes, or interests, or needs. I mean, you and I are, you know, as human beings, we do something and we get satisfaction out of it. I have a motivation for doing the research I do. And if that doesn't happen, I feel disappointed, I can feel sad.

• 30:03Speaker 01

  I also feel accountable when I put out, you know, a research paper. I know if there are errors in it, I know I'm the one responsible for it. So there is none of that. So when an AI system gives you an output, it's not because it might be worried about if it's an incorrect answer, or it's because it wants to please you. It's just a chatbot that is designed to provide, you know, given answers, again, based on prompt. So this idea of AI systems causing existential risk leans on.

• 30:41Speaker 01

  This huge leap of faith that requires you to believe that there is intention, there is emotion, there is motivation, all these human characters, all these things that makes us human, but it's just not there. It can't emerge out of nowhere. That's part of what makes us different and unique as individuals, as biological organisms. These are things that are hardwired on us. These are also things that makes us human. This is why I fret still. Of course, we have to worry about powerful people using AI to do terrible things. And what worries me is over the past year, and especially since the rise of Trump and

• 31:23Speaker 01

  since the Trump administration came to power, you have a lot of large corporations really abandoning their voluntary pledge to protect basic fundamental rights. So Meta has, for example, walked back their commitment for AI, their commitment to fact check and monitor their social media platforms. Against hate speech too as well, which is what? Exactly. What really is worrying also now is you have superpowers and powerful governments like the US government, the UK government, even the European Union itself, using AI, moving into AI for surveillance, for military purposes, for warfares. And a

• 32:08Speaker 01

  lot of AI companies, starting from, you know, OpenAI, Meta, Amazon, Google, they had voluntary principles not to use AI for military purposes, but over the past year, all of them have abandoned that. So even here across in the EU, you have a French AI company called Mistral announcing that they are open to working with European governments to provide military AI. So of course, we have to worry about governments using AI under the guise of national security, which really means, you know, monitoring and surveillance and squishing dissent. And really, this is against, you know, fundamental rights for freedom of expression, freedom

• 32:50Speaker 01

  of movement and so on. So we have to worry about AI, but AI in the hands of powerful governments and people in position of power rather than the AI itself, because the AI can't do anything by itself.

• 33:05Alie Ward?

  Is it sort of like the guns don't kill people, people kill people kind of a situation?

• 33:10Speaker 01

  Yeah, exactly.

• 33:11Alie Ward?

  How is that working out for us in the US, though? Well, death by firearm is the leading cause of mortality for teens and children in the US, according to the Pew Research Center. And over half of our nearly 50,000 gun deaths a year in the US are suicides. This is not going well. And that's because the NRA slogan, guns don't kill people, people do, is what is known as bumper sticker logic, or a misdirection, also called a false dichotomy, or plainly speaking, a fallacy, according to philosophers. So giving AI a sense of technological neutrality is a bit misguided. The regulations being walked back is terrifying,

• 33:53Alie Ward?

  especially trying to put trust in a government to stop things when a lot of our people in power don't know how to use their own printers. You know, so some of the questions in like the congressional hearings are like, how does this even work to?

• 34:08Speaker 02

  Does Google track my movement? Does Google, through this phone, know that I have moved here and moved over to the left?

• 34:17Alie Ward?

  Which is terrifying. So maybe they don't have morals and guilt and things like that and ambitions. But I was looking at some research showing that AI is being trained to become more and more sexist, more and more xenophobic, more and more racist, use more and more hate speech. And is it learning from the worst of humanity? Is it amplifying it? Is that just exposing how much hate is in the world?

• 34:48Speaker 01

  Yeah, so let's maybe walk back 20 years.

• 34:55Speaker 01

  Because that's when, you know, real progress in AI started to emerge. I mean, we've had a lot of the core principles for AI, you know, since in the 1950s, 60s, 70s.

• 35:11Speaker 01

  Eighties, like some of the foundational papers about reinforcement learning, deep learning were written in the nineteen eighties. So Jeff Hinton's famous paper on, I think it was convolutional learning, was written, you know, late nineteen eighties.

• 35:28Alie Ward?

  We don't have to go too deep into it, but I do want to tell you that Jeffrey Hinton is apparently considered the godfather of AI and a leading figure in the deep learning world. And in twenty twenty four, he won the Nobel Prize for his work. He's also worked for Google Brain and then he quit Google because he wanted to, quote, freely speak about the risks of AI. Quit Google so he could talk about it. Now, in twenty twenty three, during a CBS Saturday Morning News segment, he warned about deliberate misuse by malicious actors, unemployment and existential risk involving AI. He is very much in favor of research on the risks of what could become a monster that he helped create.

• 36:10Alie Ward?

  He's like, yo, we need some safety guidelines and regulations, buddies, and that is not really happening. But yes, he is among a few who, over the last many decades, drove these innovations.

• 36:22Speaker 01

  But what really made the AI revolution possible is the World Wide Web. With the emergence of the World Wide Web, it became possible to kind of scrape, to kind of gather, harvest massive amounts of data from the World Wide Web, you know, through chat forums or domains like Wikipedia. They are really a core element of training AI, at least for text data. So that means that a lot of our training material for AI comes from the World Wide Web, whether it's our digital traces, whether, you know, it's the pictures we put on social media, pictures of your kids, your dogs, yourselves and so on, or the kind of infrastructure, digital infrastructure,

• 37:07Speaker 01

  like Google is everywhere, has dominated, whether you want to email or, you know, prepare a presentation or write a document, Google has provided the infrastructure. That means they have the infrastructure to constantly harvest training data. This means that a lot of the data. That we are using for training reflects, you know, humanities' beauty, but also our cruelty and the ugliness of humanity.

• 37:34Alie Ward?

  And just last week, a tech report released by Google admitted that its Gemini 2.5 flash model is more likely than its predecessor model 2.0 to generate results outside of its safety guidelines. And images are even worse than text at that. And I mentioned this in a 2023 episode we did with Dr. Nita Farhany about neurotechnology. But around Juneteenth of that year, I saw this viral tweet about ChatGPT not acknowledging that the Texas and Oklahoma border, the Panhandle, was in fact influenced by Texas desiring to stay a slave state, which is fact that ChatGPT would not acknowledge. So Dr. Farhany notes that when an AI is built on racist,

• 38:17Alie Ward?

  sexist, xenophobic, et cetera, data set, the results, like history itself, are not kind to minorized identities, she says.

• 38:27Speaker 01

  It reflects, you know, societal norms. It reflects, you know, historical injustices and so on. Unless you really delve into the data set and ensure that you do a thorough job of cleaning the data set. We've audited numerous data sets and you find content that shouldn't be there. You find, you know, images of genocide, images of, you know, child rape. One of the early data sets we audited back in 2019 was a data set called 80 Million Tiny Images. It was held by MIT and we found several thousands of images,

• 39:09Speaker 01

  really problematic images, images of black people labeled with the N word, images of women labeled with the B word, the C word, and words I can't really say on air.

• 39:22Alie Ward?

  So while the upside of AI is detecting cancer from scans earlier or predicting tornado patterns, there's also so much concern. Now, Dr. Martin Luther King Jr. observed and proclaimed that the arc of the moral universe is long. But it bends toward justice. But I think we might consider that the arc of the internet is short and it bends towards smut and hate.

• 39:49Speaker 01

  So you can assume any data you collect from the web is really horrible. And in one of the recent audits, actually, we found an overwhelming amount of women, women concepts, really represented by images that come from the pornographic space. So massive amounts of the web is also really pornographic and really problematic content. So you have to do a lot of filtering. So as a result, this is why, you know, D E I initiatives, this is why obligations to audit your data set, to ensure that toxic contents have been removed and so on. This is why it's so critical.

• 40:32Alie Ward?

  So an A I is only as ethical as its data sets. And the internet is a weird, dark place where people say things they would never say in person. So the data sets are feeding that.

• 40:43Speaker 01

  But as we are seeing now, a lot of these companies are abandoning their pledges and we're really walking backward. But for any given A I system, whether it's a predictive system or classification or generating, you can assume that deeply held societal injustices and norms will be reflected in how that A I performs, in the kind of output the A I gives you. So that's the default. So we have to work backwards to ensure we are removing those biases.

• 41:13Alie Ward?

  Let's say that some of the comments online, some of the hate online is A I generated comments, which I, sometimes I'll look at now X and I'll say, who are all these people? Like, why are comments getting meaner and meaner? With Facebook with a lack of fact checking, more and more sort of hateful speech? Does that mean that the next tokens and data sets pick up on that and say, oh, this is how people think? And then the next one? So does it get amplified like mercury toxicity and like a tuna fish?

• 41:46Speaker 04

  That's one way of putting it.

• 41:47Speaker 01

  Yeah.

• 41:48Alie Ward?

  Okay.

• 41:49Speaker 01

  Yes, yes, you are encoding those biases and you are exaggerating them.

• 41:54Alie Ward?

  Uh huh, yeah.

• 41:55Speaker 01

  The technical drawback is that, so we train a given AI for a next word prediction. For example, it's based on, you know, these massive amounts of data that kind of tells you how people text, how people use language, for English, for example, how people, you know, construct a coherent sentence. That data, that training data comes from actual people activities, people interactions.

• 42:20Alie Ward?

  Uh huh.

• 42:21Speaker 01

  That is your baseline, so to speak, in when you are modeling how, you know, language operates. But now, as you said, as the World Wide Web is filled more and more with, you know, synthetic text or syntax data that comes from generative AI system itself, then your AI system has no frame of reference. It tends to forget, so the quality of the output starts to deteriorate.

• 42:49Alie Ward?

  Oh, that's so scary.

• 42:52Speaker 01

  So this is called model collapse.

• 42:54Alie Ward?

  Okay, does this keep you up at night? Does it? I mean, I know that it's like, don't be afraid, don't be afraid, but it's also like, this is very new territory for humanity, right?

• 43:06Speaker 01

  Yeah, but at the end of the day, I mean, people should be in control. If an AI system starts producing outputs that is rubbish, that is irrelevant, I don't think it should scare us. It should make people like, okay, that's not helpful to me anymore, so I'm not gonna use it.

• 43:22Alie Ward?

  Maybe the more it unravels and crashes out, the less people will rely on it. But of course, that hinges on being able to tell that it's spitting nonsense at you. And in this day and age, the world is so profoundly absurd that truly anything is believable. And Dr. Brahani says that public education is key and just getting the word out that a lot of what we think about AI's capabilities are just big corporations. Pumping out hype and PR. But the auditors on the inside, like her and her lab, know that, woo, boy, howdy, hot damn. It is a bunch of horse bucky flim flam and not to believe the hype.

• 44:00Speaker 01

  The actual performance is nowhere near what the developers claim. So these are the facts that we really have to communicate. A lot of the AI systems that we are interacting with are actually subpar in terms of performance, in terms of what they are supposed to do, in terms of what people expect them to do. Because these big corporations have really mastered public communication and PR, a lot of like the failures or the drawbacks of AI systems are new to people when you actually communicate it. But this should really be like common knowledge. And if people want to use AI, they should know both the strengths and what they can do with it, but also where the limitations are and what it can't do for them.

• 44:46Alie Ward?

  Does abstention work? Does not going on meta and giving them more fodder? Does not using chat GPT? Does any sort of like boycott work?

• 44:59Speaker 01

  Yes and no. On the one hand, so a lot of these AI systems have really cleverly been integrated into the social infrastructure.

• 45:12Alie Ward?

  Yeah.

• 45:12Speaker 01

  So for example, I'm not on Facebook. I haven't been on Facebook for over 10 years. But the apartment complex I live in can only be communicated via Facebook groups.

• 45:22Alie Ward?

  Yeah.

• 45:22Speaker 01

  I still refuse to create a Facebook account. But situations like this really gives you very little option to abstain, to not use these platforms. And you can't avoid Google, for example.

• 45:35Alie Ward?

  Yeah.

• 45:35Speaker 01

  Google search and like, you know, Gmail and like Google Docs. It makes it really difficult. If you want to apply for a job, almost all companies now use some kind of AI to filter your CV before it reaches human. So in some. Senses, you don't even have the option to opt out. If you are, you know, someone looking for a job, you can't say, "Oh, I don't want you to use an AI system to sift through my CV."

• 46:05Alie Ward?

  Yeah.

• 46:05Speaker 01

  It's just like.

• 46:06Alie Ward?

  It's gonna happen.

• 46:07Speaker 01

  Yeah, it's gonna happen.

• 46:09Alie Ward?

  Dr. Verhany says that it's pretty unavoidable. And I have asked tech lawyers, and even they don't read Apple's terms and conditions. They're like, "I just checked the box."

• 46:19Speaker 01

  So instead of using WhatsApp, which is owned by Meta, which, you know, really gathers all your text, all your information, we can move to other messaging apps like Signal. So Signal has, you know, end-to-end encryption. There is no backdoor. Nobody can access it, not even governments. This is one of the things Meredith Whitaker, the CEO of Signal, has been really strong in standing up to large governments is that nobody should have a backend access. That gives them the opportunity to gather data.

• 46:50Alie Ward?

  And yes, Signal is run by a nonprofit foundation, SignalFoundation.org. And Meredith Whitaker is Signal's president, and she had worked at Google for 10 years. And she was raising concerns about their AI. And she was also a core organizer of the 2018 Google walkout in protest of sexual harassment there and pay inequities. She also advises government agencies on AI safety and privacy laws. So Signal, good. Yay, Signal. And many recently laid off government staff that I know of will only communicate via Signal, which is kind of telling in terms of their own safety concerns. But yes, use Signal.

• 47:26Speaker 01

  So we can do some things. We can use less and less of these large corporations' infrastructure. And we can use, you know, more open source tools. But also sometimes, you know, it's just out of your control. But every little helps and every awareness, you know, it kind of culminates and it will eventually lead to, you know, this massive switch, I hope, at least.

• 47:49Alie Ward?

  That's encouraging.

• 47:50Speaker 01

  I hope. Yeah.

• 47:51Alie Ward?

  And can I ask you some questions from listeners? Is that okay?

• 47:55Speaker 01

  Yeah, for sure.

• 47:56Alie Ward?

  But before we do, let's give away some money. And this week, Dr. Bahani selected the cause, the municipality of Gaza and UNRWA, which directly supports Palestine refugees and displaced families in Gaza. They say every donation, no matter the amount, helps them reach families with life-saving food assistance, shelter, healthcare, and more. And for more info, you can see donate.unrwa.org, which is linked in the show notes. And for more on the ongoing humanitarian crisis in Gaza, please see our Genocideology episode with global expert in crimes of atrocity, Dr. Dirk Moses, which we will also link in the show notes. So a quick break now. This episode is sponsored by Seed. Bloating, irregular bowel movement, gas.

• 48:41Alie Ward?

  When your gut is off, it gets your attention. Seed's DS1 is a three-time award-winning probiotic, clinically shown to improve regularity, reduce bloating, and alleviate gas so you can move and eat freely. Take better care of your gut with Seed. Get 25% off with code 25OLOGIES at seed.com/ologies. These statements have not been evaluated by the FDA. These products are not intended to diagnose, treat, cure, or prevent any disease. We have to say that. You know, it's fall, it's back to your routines. But just because you're not half naked next to a swimming pool doesn't mean you don't want to look your best when you leave the house. I also don't like to spend time doing these things. So I like minimalist beauty products, and I like good stuff.

• 49:25Alie Ward?

  And Merit fits the bill. I've been using them recently. Love them. So they make a bunch of great stuff. And I like to start with the Great Skin Serum. It feels so good and so soft on your face. Also, I have the Flush Balm that I love. It's this little, easy-to-use blush. They also have something called the Minimalist, which works as a foundation and concealer, which is great for people who love a BOGO. And I also love that Merit's products are clean, they're vegan, they're cruelty-free, and their skincare ingredients do leave your face looking glowy and lovely. And fall, it's a perfect time for a fresh start. You can make it your back-to-me season with Merit. And right now, Merit Beauty is offering our listeners their signature makeup bag, which, if you've seen it, it's not just a makeup bag, but really like an accessory,

• 50:10Alie Ward?

  free with your first order at meritbeauty.com. That is meritbeauty, M-E-R-I-T, beauty.com. Meritbeauty.com. It's on my face right now. This podcast is brought to you by Squarespace. If you haven't heard me gushing about Squarespace for years, it's an all-in-one website platform. Whether you're trying to grow a business you have, or if you're just a baby business getting started, it has everything you need. That's where I secured my domain name. It helped me build a professional site. I can update it so easily. I've been using Squarespace since before Ologies existed. After procrastinating for years, I literally built my website in one evening. They have templates, they have flexible editing tools. Squarespace also makes it easy to share your work.

• 50:55Alie Ward?

  You can book clients, you can get paid. They have built-in tools for scheduling and invoicing and email, all in one place. Whenever someone I know needs a website, whether they're a scientist that needs to put their work up, or someone who's just starting a business, I'm like, dude, Squarespace. So head to Squarespace.com/Ologies for a free trial. And when you're ready to launch, use offer code OLOGIES to save 10% off your first purchase of a website or domain. You can do it. Being safer on the road starts with the person behind the wheel. That's Mazda's philosophy when designing every vehicle. Their proactive safety features help keep you more aware, connected, and engaged during every drive, which means your car does more than help protect you.

• 51:37Alie Ward?

  It makes you a safer driver, which helps protect everyone. That's why Mazda has been named Consumer Reports' safest new car brand. To learn more, visit MazdaUSA.com. Consumer Reports does not endorse products or services. Every time I record a Quince ad, I'm like, look at me, wearing head-to-toe Quince. I feel like this is enough of an ad. I can just stop here. But if you're like, huh, what is that? Quince sells clothes and shoes and bags, even stuff for your house that are classic, well-made, and priced at 50% to 80% less than similar brands because they've cut out the middleman. They work directly with ethical factories. So you have luxury quality pieces without a luxury price tag. Perhaps this holiday season, do you know what it feels like to get a cashmere sweater as a gift?

• 52:22Alie Ward?

  I don't, because I'm the one getting them for people. And so can you, and they're affordable, too. They also have 14 karat gold earrings, high-quality down outerwear. Cozy gloves made of cashmere leather bags. They have great bedding, cookware luggage. So whether you're giving that to you or to someone else, you can give the gift of quality this holiday season. Download the Quince app for app exclusive offers, or you can go to quince.com. Get free shipping on your order and 365 day returns, now available in Canada and the UK too. And when Quince asks you where you heard about them, just let them know it was Ologies. It's the best way to support the show. Okay, we are back. Let's run through some questions

• 53:06Alie Ward?

  from your real squishy brains made of human beings out there. There's some great ones.

• 53:13Alie Ward?

  Job displacement. Carla De Azivo, Elia Myers, Red Tongue, Jennifer Grogan, Ian, Jenna Congdon, Rosa, Rebecca Rome, Other Maya, Sam Nelson, Howard Nunes. All these people wanted to know, in Ian's words, will all jobs be obsolete soon? Do the people working on AI give any thought to compensating people for the lost income? Jenna Congdon said, when will AI get so good that human writers are basically crowded out of a job? This goes for visual art as well. In a capitalist economy, when you've got to hustle to make money as it is, what is going to happen job wise? Do you think? Or do AI experts such as yourself?

• 53:53Speaker 01

  So some of the worry about job displacement is genuine and grounded on, you know, real worry. You hear even the so-called good fathers saying things like, you shouldn't bother learning code or like the job of software engineering, for example, will become obsolete and so on.

• 54:15Alie Ward?

  So whether you're a software developer or a writer or an artist, Dr. Brahani says.

• 54:20Speaker 01

  I don't think AI will fully automate. AI will fully replace human task force. Because at the end of the day. What even the most advanced AI systems do is really kind of aggregate information and kind of outputs a very mediocre, whether it's image or text.

• 54:46Alie Ward?

  Some of them are so good, though. Some of the art is so good and you're like.

• 54:49Speaker 01

  But some of the arts, it's not just the pure, the raw outputs. People have tweaked probably like a thousand times. People have tweaked it. People have spent hours perfecting the right prompt and so on. So there is always people in the loop. There is always, whether it's data preparing, you know, data annotation, data curation to building the AI system itself, to then kind of ensuring the output is something appealing. You really, you need people through and through.

• 55:22Alie Ward?

  So for me, as a former newspaper journalist, and I was also a newspaper illustrator, I, I'm not as optimistic. So, so many writers are copywriters who are making content and articles for websites to raise their profile. And now I'm hearing from those people that articles are just written by AI and they are full of shit. And just doing this aside is making me depressed and my chest hurts. But Dr. Brahani is an expert, so I'm going to try to find some bright spots. And before she had mentioned that lawsuit with OpenAI and the New York Times. And I was looking for it and I found a recent article. This was published literally yesterday, which had the headline, "AI is Getting More Powerful, But Its Hallucinations Are Getting Worse.

• 56:06Alie Ward?

  A New Wave of Reasoning Systems From Companies Like OpenAI Is Producing Incorrect Information More Often. Even The Companies Don't Know Why." That's the headline. And this New York Times article explained that AI systems do not and cannot decide what is true and what is false. And sometimes they just make stuff up, a phenomenon that some AI researchers call hallucinations.

• 56:30Alie Ward?

  And in one test, the article says hallucination rates of newer AI systems were as high as 79%. And I also want to note that my spell check tried to get me to change the it's in the headline to one with an apostrophe, which is incorrect. So, computers, what's going on? But yes, Dr. Bahani says that a lot of journalism has been replaced by AI, even though we all know that the generative system is unreliable.

• 56:57Speaker 01

  It hallucinates. A lot of the time, it gives you information that sounds coherent, that seems factual, but it's just absolutely made up. It just, there is, it even sometimes gives you citations and so on of things that don't exist. So, we always need people to babysit AI, so to speak. So, a writer might be, you know, your hours might be reduced and you might be getting paid less and your company might be bringing AI to kind of do the bulk of the job, but still you can't put out the raw outputs because most of the time, it's not even, you know, legible.

• 57:37Speaker 01

  So, the role of writers and artists and journalists and so on becomes more of kind of a babysitter for AI, verifying the information that's been put out, kind of ensuring it makes sense and so on.

• 57:51Speaker 05

  That's right, Kenny. The babysitter is dead.

• 57:53Speaker 01

  To some extent, the answer is yes and no. Humans will always remain at the heart of AI. The minute human involvement ceases, the minute AI stops operating. Because AI is human through and through, as I said, you know, from the data that's gathered from humans and so much work goes into data preparation, data annotation, cleaning up the data, detoxifying the data. And unfortunately, a lot of these tasks are located to, you know, the developing world. So, you have a lot of data workers in Kenya, in Nigeria, in Ethiopia, in India, for example. That really do the dirty work of AI. There are even a bunch of stories where you have Amazon checkout, for example, AI checkout

• 58:36Speaker 01

  or self-checkout, where Amazon was introducing this AI where you can just collect groceries and your items and just walk out and the AI is supposed to kind of identify what you have picked up and charge you from your credit card for whatever you have used. But then it turns out that it was actually data workers in India that were scanning every item you are picking up, so.

• 59:02Alie Ward?

  Oh, man. What a world.

• 59:06Speaker 01

  Yeah. Yeah. And I mean, McDonald's also recently partnered with IBM or one of those companies to have like an AI drive-through where AI systems take order. And they have to close it within the next few weeks because people were getting orders of like, you know, bacon on top of ice cream and things like that.

• 59:27Speaker 06

  You added bacon to my ice cream. I don't want bacon.

• 59:32Alie Ward?

  Why raise the national minimum wage for the first time since 2009 when you can just spend billions of dollars tweaking unpaid machines? Like, welcome to the future, maybe.

• 59:42Speaker 01

  So I guess the point I'm trying to make is like you always need humans for AI to to function and operate as it's supposed to. Because at the end of this day, these are like really mere machines that don't have, you know, intention, understanding, motivation, and so on, like we humans are.

• 1:00:01Alie Ward?

  So maybe our jobs will look different, but there will be jobs.

• 1:00:04Speaker 01

  Yes.

• 1:00:05Alie Ward?

  I know a lot of people, myself included, wanted to know the environmental impact. Lily, a bunch of folks. And first time question askers Eleanor Bundy and Megan M. And we also did a recent episode for Earth Day with this climate activist and humanitarian rights lawyer, Adam Met, who said that AI could be solving some environmental concerns, which is optimistic. But what does an AI expert's take on that? Megan Walker asked environmentally, how bad is AI when compared to the current computing we do? Yeah, what's going on?

• 1:00:33Speaker 01

  Yeah, yeah, yeah.

• 1:00:34Alie Ward?

  How much energy does it use?

• 1:00:37Speaker 01

  Yeah, so again, like, we have very little information about training data. The kind of energy consumption used by AI systems is very opaque. There is very little transparency.

• 1:00:51Alie Ward?

  Okay, but what is the damage generally?

• 1:00:53Speaker 01

  So generative AI really consumes massive amount of power compared to traditional AI. For example, if you are using Google to put a prompt, say, you know, how many glasses of water should I drink per day? And if you do the exact same prompt and you ask the generative system such as ChatGPT, people estimate you use about 10 times more energy to process that query and to generate answers.

• 1:01:20Alie Ward?

  I wanted to go straight to the source. So I used Google AI and ChatGPT for the first time, asking them both, how much energy does ChatGPT use as opposed to Google? Now, Google AI said in what I hope is a snotty tone that ChatGPT consumes significantly more energy per query, five to 10 times more electricity than a standard Google search. Now, it cited a 2024 Goldman Sachs report titled, "AI is poised to drive 160% increase in data center power demand." Then I asked ChatGPT, and it said that its version four can use up to 30 times more energy than a basic Google search. And it also noted, I like to think defensively,

• 1:02:03Alie Ward?

  that Google has had decades to optimize for a lower footprint. Now, Dr. Brahani says that the energy consumption of generative AI systems has become indeed a big issue, and that in countries like Ireland, the data centers are power hogs.

• 1:02:19Speaker 01

  The compute resources required to running AI systems equals or is more than the total amount of energy that's required to run Irish households. But in places like Texas, Sometimes that energy consumption is taken away from households to run data centers.

• 1:02:41Alie Ward?

  Wow.

• 1:02:42Speaker 01

  So it kind of results in reduced energy for households. And this is before we even get into the massive tons of water that you need to cool down data centers.

• 1:02:56Alie Ward?

  Oh.

• 1:02:56Speaker 01

  Yeah.

• 1:02:57Alie Ward?

  I didn't even think about that.

• 1:02:58Speaker 01

  Yeah, and and the water also has to be pure because you can't use, say, you know, ocean water or seawater because of the sea salt that might damage the servers and so on. So again, there is competition. It tends to be when you use water that is used for households. As a result, you know, people tend to to pay for the consequences of that. So yeah, water consumption is another massive area as well.

• 1:03:24Alie Ward?

  And do you think that more companies will look toward some sort of nuclear power for their supercomputers? Or is that still too highly regulated?

• 1:03:34Speaker 01

  Uh, I think companies like Google are actually talking about using nuclear power. But yes, that option is being considered. Yeah.

• 1:03:44Alie Ward?

  How about healthcare? Several people, Benjamin, Brenna Schwert, Annalise De Young, Emil, Nikki G. asks, how can AI be ethically applied to healthcare? Like data analytics, treatment options, medical imaging interpretation, second opinions? Is there some hope there for it?

• 1:04:01Speaker 01

  Yeah, I think there is some hope. There is some hope for sure. I think there is some hope in, you know, numerous domains for AI to be useful.

• 1:04:11Alie Ward?

  Okay.

• 1:04:12Speaker 01

  However, that just remains a theory. It's it's possible in theory.

• 1:04:16Alie Ward?

  Oh.

• 1:04:17Speaker 01

  But the problem, there are a bunch of problems. One of them is that generative systems are fundamentally unreliable. So for example, there is a new audit that came out, I think, towards the end of January, where they looked at this new AI tool where the system kind of records your conversations between, say, a healthcare provider and a patient. And it summarizes the conversation. And it kind of, it's supposed to reduce a lot of the work for nurses and so on. And what they found was that in some cases, eight out of the 10 summaries were hallucinations.

• 1:05:00Speaker 01

  No.

• 1:05:02Speaker 01

  So generative systems tend to be unreliable. And the other thing is because a lot of these tools that are supposed to be used in healthcare tend to be built by businesses with the objective of maximizing profits. They tend to have a different kind of objective than say, you know, what's good for the patient. So another famous case is United Health that is in court at the moment where they were using a suit of about fifty algorithms to look at mental health services. And what they found was that they were using, you know, cost as a proxy rather than the need of the patient as a proxy.

• 1:05:45Speaker 01

  And they were cutting a lot of services, a lot of like, you know, therapy services and meditations and other necessary services. Again, because they are, you know, they are looking at the wrong motivation, the wrong proxy. They are looking to save the company money rather than to ground what they do in the needs of the patient. So if we correct, for example, hallucinations and biases in AI systems, and if we kind of, it's impossible to strip down all, you know, capitalist motivations. But if, you know, capitalist motivations come second to, you know, the, the needs of the patient, then it's possible to kind of develop AI systems in various areas of

• 1:06:30Speaker 01

  healthcare that prioritize patients, that prioritize people as opposed to just, you know, inserting technology for the sake of having technology.

• 1:06:40Alie Ward?

  Yeah.

• 1:06:40Speaker 01

  And also for using technology to maximize profit as opposed to ensuring patient safety.

• 1:06:46Alie Ward?

  Which is, once again, good luck. I mean, our healthcare is above and beyond frustrating. But a lot of people wanted to know, Samwise, Emily Heard, Amalia, Magda Kasauka, wanted to know, Kira Henningston asked, first time question asker, how do we feel about AI and chatbots and using them in high schools? Schoolwork? Samwise, AI use in schools? Thoughts? Is there a way to flag it? Are we doing education and injustice?

• 1:07:15Speaker 01

  Hmm.

• 1:07:17Speaker 01

  Yeah, so on the one hand, I know some people that find using AI chatbots really helpful. You give it a prompt, it gives you just a bunch of answers. Of course, these are people that know how to craft the perfect prompt, that know where AI can be useful and where it might fail you. So with all that in mind, it can be useful, but you need to be an expert. Having said that, for young kids, studies are starting to emerge that, for example, they did a control study, I think with over 3,000 students,

• 1:07:59Speaker 01

  where some of the students were given chatbots to help them with, I think it's maths problems, the others weren't, and they did a test. So what they found was that the kids that had chatbots did better than the kids that didn't have. Then they performed another test a few weeks later, and they found that the kids that used chatbots performed way worse than the kids that didn't have. So people are realizing that these systems inhibit learning. Of course, you know, education is not just information dissemination, the teacher going into class and just like telling the students facts,

• 1:08:43Speaker 01

  but rather it's an interaction, it's a two-way street, both for the student and the teacher, developing the skills, especially critical skills, to analyze and to decipher fact from fiction, you know, information from misinformation and so on. And when you use AI chatbots without knowing their limitations, you tend to kind of trust the output, you tend to treat it as fact, but also it inhibits your learning, it inhibits your critical skills. And if you don't have the knowledge to begin with to verify the answer, you have no way of knowing what you have, what you are getting is correct or incorrect. So in the long term, studies are coming out to show that

• 1:09:27Speaker 01

  they might seem helpful in the immediate term, but in the long term, these chatbots might be inhibiting the learning process.

• 1:09:36Alie Ward?

  Last listener question, and I know my husband has this question, too. Deviency, Shara Rempel, and Chelsea and her doctor Charlie want to know. Chelsea asked, why does AI do better when you threaten it? Is that ethical? Because it doesn't feel like a good precedent to set in any part of life. Deviency asked, is it weird that I feel the need to say please and thank you when talking to chatbots? Will the AI overlords be nicer to me when they take over? I assume all of my conversations will be logged for eternity. And Jarrett, my husband, also doesn't use ChatGPT very much. But when it came out, he was trying to teach it to be civil. And I was like, boy, I don't think that's going to work.

• 1:10:14Speaker 01

  Yeah.

• 1:10:14Alie Ward?

  Do manners matter?

• 1:10:16Speaker 01

  Yeah, yeah, yeah, yeah. So for models like ChatGPT, they have what they call a knowledge cut of date. So the training data, the, you know, your interaction won't really input into into the learning system of the the model. It's the training data set for, I think, ChatGPT that ends, I think, around 2021 or 2022. So ChatGPT, for example, can't give you a coherent answer for any event that has happened recently.

• 1:10:41Alie Ward?

  So different models are using data from different timelines. They have to collect it and clean it and process it first. So it's not as real time as I thought it was, or as some people might expect.

• 1:10:51Speaker 01

  For I for kind of when you speak to it, was it aggressively threatening?

• 1:10:57Alie Ward?

  Yeah, threateningly. Why does why does it do better when you threaten it?

• 1:11:01Speaker 01

  So it's the first time I'm hearing this. So I should I should try it out. I should check it out and see if that also happens to me. But yeah, it's it's the first time I'm hearing it.

• 1:11:11Alie Ward?

  OK, let's hit the books for this. Specifically, a 2024 study about how to interact with large language models. And it's titled, Should We Respect LLMs? A Cross-Lingual Study on the Influence of Prompt Politeness on LLM Performance. So this abstract explains that they did testing in English, Japanese, and Chinese language models, and that as the politeness level descends, the answers generated get shorter.

• 1:11:38Alie Ward?

  However, on the far side of the rude scale, impolite prompts often result in poor performance, but overly polite language does not guarantee better outcomes. And the best politeness level is different according to the language. And they say that this suggests that LLMs not only reflect human behavior, but are also influenced by language, particularly in different cultural contexts. So what about the future? The researchers say that it is conjectured that GPT-4, being a superior model, might prioritize the task itself and effectively control its tendency to, quote, "argue at a low politeness level." So as it matures, it just won't engage.

• 1:12:21Alie Ward?

  It's like this AI new generation has been to therapy, if it were a person, which it's not.

• 1:12:27Speaker 01

  And as to the AI overlords, again, I mean, it's just models. It's just, you know, data sets and algorithms and, you know, connection of networks.

• 1:12:37Alie Ward?

  Okay.

• 1:12:38Speaker 01

  There is no kind of all-knowing, godlike, all-seeing AI. But of course, you know, the people that are running AI companies come close to that because they have access to their data, because they have access to the algorithms. So you might worry about those people, you know, using your data. It has like almost invisible but very nuanced downstream impact.

• 1:13:03Alie Ward?

  Oh.

• 1:13:04Speaker 01

  So in the US, for example, authorities are forcing companies like Meta to give up data so that authorities are hunting down women that had abortions, for example, in areas where abortion is prohibited. Law enforcement is working with Amazon, for example, for, was it Amazon Ring?

• 1:13:26Alie Ward?

  Oh, right. This is that camera-enabled and Amazon-owned Ring doorbell.

• 1:13:32Speaker 01

  So law enforcement, like ICE.

• 1:13:36Speaker 01

  Uses that kind of data from those to to do something to even deport people. So what we should worry about is not really AI overlords, but this company is working with powerful entities to really kind of identify people that might be in trouble with the law or that might be doing, you know, that that violates the law because that data gives them access, gives them the knowledge about the whereabouts and the interactions and the activities of people.

• 1:14:06Alie Ward?

  Yeah. And previously, according to a 2020 Newsweek article titled, "Police are monitoring Black Lives Matter protests with Ring doorbell data and drones," activists say. It's reported that Amazon Ring has video sharing partnerships with more than 1,300 law enforcement agencies across the US. However, in January 2024, Ring said that it would stop letting police departments request and receive users' footage on its app. Now, on the flip side, some Ring doorbell owners are posting on the Ring Neighbors app when ICE raids are going down locally and they're alerting their community. Now, Ring, of course, notes that those are user-generated posts. It has nothing to do with them. Whether or not they'll censor those user-generated posts is like anyone's guess.

• 1:14:51Alie Ward?

  Hey, let's take a welcome departure from reality for a sec, shall we? Do any movies get it right? Like, does The Matrix get it right? Any, does AI, that old Spielberg movie? Does anyone actually get AI right? Or does that drive you absolutely insane to watch TV?

• 1:15:06Speaker 01

  So I love science fiction, actually. Like, The Matrix is one of my favorite movies.

• 1:15:10Alie Ward?

  I knew it, I knew it. It's a good one.

• 1:15:12Speaker 01

  But also, that's like, that's nowhere. You have to treat science fiction as science fiction for the sake of like, it's really, some really good science fiction really brings you into a world that you couldn't even envision. So I love that element about science fiction. But a lot of this, like, robot uprising, Terminator-like, movies are really just for entertainment. There is nothing that can be extrapolated and said, oh, this could happen to real AI. But you have kind of very nuanced sci-fi movies that nail it. So you have Continuum. It's not a movie, it's a series. It was on Netflix a while back, a few years back.

• 1:15:55Speaker 07

  For what? For what you're going to do.

• 1:15:57Speaker 01

  So what happens here is that, you know, as AI companies become powerful, they take over government and they become, you know, the bodies that really govern society. So that kind of sci fi is very close to reality than, you know, Terminator like movies. Yeah.

• 1:16:16Alie Ward?

  How about Black Mirror?

• 1:16:18Speaker 01

  Oh, Black Mirror is so good. I mean, Black Mirror, there are some things that are like, huh. Now, when Black Mirror came out, it was just like, wow, this could happen. And now it's like, oh, that has happened. Or it's like, oh, yeah, this is, you know, this is what's happening with this and that government. So, yeah.

• 1:16:36Alie Ward?

  Wow. Yep. And the last two questions that I always ask are always worst and best. I guess your, your most loathed thing about AI and your favorite thing about it. I guess we've talked a lot about cautionary, but like in terms of what you do or in terms of your job, worst and best thing?

• 1:16:53Speaker 01

  Yeah, so the worst thing is really just the hype. As a researcher, I have my own research agenda, but the hype is so destructive. You see something that's not true being disseminated, going viral. And, you know, as an expert, it's really troubling. So you have to stop what you are doing and do some work to kind of correct it or at least attempt. So, yeah, a lot of the hype really is what really gets on my nerves. And it, it becomes also a problem in terms of getting my own work done. But what excites me about AI is I'm still extremely optimistic about AI.

• 1:17:32Speaker 01

  But unfortunately, a lot of the AI I get excited about is not something that results in, you know, massive profits. So, you know, using AI for disaster mapping, using AI for soil health monitoring and so on. These are things that really excite me, but there is no monetary value in developing AI for, for this system. So these are the things that really get me excited, that really make me feel like, wow, this is powerful tool that we can use to actually do some good in the world.

• 1:18:04Alie Ward?

  Yeah, we could make sure that everyone is fed and has healthcare and that resources are allocated in a way that's fair. And we just don't because of money.

• 1:18:13Speaker 01

  Yeah, because it doesn't make you money.

• 1:18:15Alie Ward?

  Yeah, which is, I think, once again, money is the root of all.

• 1:18:18Speaker 01

  Yes.

• 1:18:19Alie Ward?

  Yeah, yeah, yeah. Um, thank you so much for doing this. This has been so illuminating and it's great to talk to someone who knows their shit about this.

• 1:18:27Speaker 01

  Thank you so much for having me. I really enjoyed our conversations.

• 1:18:32Alie Ward?

  So ask real people, real not smart and important questions because how else are we supposed to learn anything? So thank you so much to Trinity College's Dr. Brahani for sitting down with me and making the trip to Ireland so eventful. I loved this talk. And you can find links to her and her work in the show notes as well as to the cause of the week. We are at Ologies on Meta-owned Instagram and on Blue Sky. And I'm giving my data as Ally Ward with just one L on both. And our website has links to all the studies we talked about and that link is in the show notes. If you're looking to become a patron, you can go to patreon.com/ologies and you can join up there. If you need shorter, kid-friendly versions of Ologies episodes, we have them for free in their own feed. Just look for Smallologies.

• 1:19:16Alie Ward?

  That's also linked to the show notes. Please spread the word on that. And we have Ologies merch at ologiesmerch.com. Thank you to Erin Talbert for adminning the Ologies Podcast Facebook group. Aveline Malik does our professional, human-made transcripts. Kelly Ardware is the website. Noelle Dilworth is our flesh and blood scheduling producer. Human organism Susan Hale managing directs the whole show. A live editor, Jake Chaffee, helps put it together. And the connective tissue lead editor is Mercedes Maitland of Maitland Audio. And Nick Thorburn made the theme music using his brain and ears and fingers. And if you stick around to the end of the show, I tell you a secret. And this week, it's two. One is that I think I'm going to be shooting something next week and I want to tell patrons about it first, but also I'll do some posting on social media if and when it happens.

• 1:20:01Alie Ward?

  I'm really excited. I don't mean to be secretive, but just send good vibes next week. I'll tell you as a secret after it happens. And the other secret is before I went to Ireland, I got a couple of those film cameras dispos because it's like, ooh, what is this? There's film in this. And I took all the pictures and I haven't gotten them developed yet. And I kind of feel like the longer you wait to get them developed, the more you'll like them. And so I don't know what the appropriate amount of time is to forget about this disposable camera and then get it developed if it should be like. A couple more months, or if I should get it developed in a year. And so now I just have this disposable camera in my backpack. And I don't know how long I should. I also don't know where to get it developed, if I'm being honest.

• 1:20:45Alie Ward?

  But anyway, if anyone has thoughts about that, feel free to advise me. That is a very analog update here for me. All right, fucking please do not use ChatGPT to write papers or illustrate anything important. Hire an illustrator if you can. Illustrators, writers, artists, musicians, please let them live. They are alive. Okay, be good. Bye bye. Pachydermatology, mammalogy, cryptozoology, lithology, nanotechnology, meteorology, geology, nephology, seriology, histology.

• 1:21:27Speaker 04

  We marveled at our own magnificence as we gave birth to AI.

• 1:21:35Speaker 08

  The Disney Plus, Hulu, HBO Max bundle. The best in entertainment in one bundle.

• 1:21:39Speaker 09

  This is gonna be a good time.

• 1:21:41Speaker 08

  With new seasons of your favorites like Percy Jackson and the Olympians on Disney Plus.

• 1:21:45Speaker 05

  Do you ever get used to all this? Nope.

• 1:21:47Speaker 08

  The scandalous All's Fair on Hulu.

• 1:21:50Speaker 05

  Let's give these girls a show.

• 1:21:51Speaker 08

  And the high society drama of the HBO original, The Gilded Age on HBO Max. What have I gotten myself into? Streaming this month. Plans starting at $21.99 a month. Terms apply. Visit DisneyPlusHuluMaxBundle.com for details.

• 1:22:05Speaker 09

  Home is for impressing, but also resting and entertaining guests in and festing and feasting and sleeping and greeting and rising and baking and making space for friends in, burning some sage in, decorating and staging. This season, visit HomeSense to get all the things that make your home, home. HomeSense, abundantly affordable.