Tech Brew Ride Home - (Bonus) A Human Algorithm With Flynn Coleman

Episode Date: December 8, 2019

I’ve been trying to read books about AI lately to get a firmer grasp on this important topic, and the best book I’ve read for both a basic grounding of the history and the state of play of AI, but also looking at potential frameworks for the technology, both ethically and socially is A Human Algorithm: How Artificial Intelligence is Redefining Who We Are by Flynn Coleman. Sponsors: PixelUnion Mealime Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to another weekend bonus episode of the Tech Meme Ride Home. I'm Brian McCullough. I've been trying to read books lately about AI to get a firmer grasp on this obviously important topic. And the best book I've read for both a basic grounding of the history and the state of play of AI, but also for looking at potential frameworks for the technology, both ethically and socially, is a human algorithm, how artificial intelligence is redefining who we are. by Flynn Coleman. We're going to talk to Flynn today. Please enjoy this great discussion about a topic that I'm still wrapping my head around. So, Flynn, I know that this is the most cliche question to ask someone who's written a book. But because, you know, I've read the book and there's obvious passion behind it, I'm curious why this topic and why this book is this,

this, why is this something that you felt was important to talk about right now? Yeah, I've been working at the intersection of human rights and technology for my entire career, so constantly thinking about questions like, what does it mean to be human, and how are we leveraging technology for the betterment of society and for the common good? And as you just mentioned, and of course, also working in this space, we are at a crossroads. We own exactly know, of course, what the future holds. But in terms of technology and specifically artificial intelligence, what's happening is we've always had new technology and tools, but we've often had more time to assimilate to these new tools and technologies, and now it's advancing

so rapidly that human beings aren't going to get that much time to assimilate. So we really have a lot of questions we need to be thinking about and conversations we need to be having on a societal level if we don't want this to get away from us and if we want to make decisions that are not just about building better and smarter tools, but how we can use them to make the world a better place. I don't know if you saw it, but this came across my transom just today that was in an ex-googler, somebody involved in AI was like, oh, we've got this limited window to control AI before it takes over. And there's lots of doom and gloom stuff about that, like Elon Musk and stuff. or whatever. You're not doom and gloom, but, and in fact, I'd say that you're mostly hopeful

about the future of AI and human beings, but it's still, you think, a really crucial time to start making decisions to make sure that the doom and gloom doesn't happen. Part of why a human algorithm is still rather radical is because ultimately it is hopeful and optimistic book. That being said, as an international human rights lawyer and a war crimes and genified lawyer, I have, of course, seen the worst of humanity, but it also means I've seen the best. So I think the short answer to your question and to all of these questions and the headlines we're seeing is, of course, we don't know. And that's actually a big premise of my book, which is that no one of us has all the answers. And we can only go so far in terms of

predicting what's going to happen. So for me, the question is less about predicting the future and more about being ready no matter what and starting to have these conversations about ethics and morals and values and compassion and who we want to be as a society. Because for me, that means that no matter what comes and we don't know what's coming, we can become better people in the process no matter what happens. Another question about where we are on the timeline of this. I talk about on the show a lot, I feel like for the general populace, the technology in general, But in specific things like the smartphone and social media, kind of just happened in the blink of an eye. One day, the tech was something that you may be used at work, but now it's in every crevice of our lives for good and bad. In terms of the timeline, do you think we're on a similar cusp with AI where one day we might blink and wake up and suddenly it will have transformed everything and be running everything?

I think that it might feel like that. It's only natural that it might feel like that. But artificial intelligence, the term has been around since the 1950s. Alan Turing, before there was even the actual technology to build his idea of a thinking machine, has been around a long time. So it's new technologies, but in essence, these are the oldest questions we've ever asked ourselves. So I think that what we need to understand is that it's moving very fast. We don't know which road it's going to take.

And I do think it's accelerated in the sense that when you're building something that's potentially smarter than humans are or could ever be, you're in a little bit of a different ballgame. That being said, history has tons of examples of technologies that we've struggled to assimilate to and then became, you know, really important parts of our lives. I think that it might feel like the blink of an eye also, because one of the things I talk about in the book is that we need a much wider public literacy and conversation and open debate about these ideas. I think some of the most dangerous threats are the most insidious, like potentially

not sharing the same reality and truth becoming fractured when we're all able to live in our own media or social media bubbles. And so that means the technology incrementally is always how it goes. So it's actually not just there one day, but one of the issues that there's so few of us involved and having a say in how we build these tools. So it could very well feel one day like suddenly it's here when, of course, there were many opportunities to weigh in along the way. Well, one of the reasons why I really enjoyed your book is that aside from the compelling arguments that you make, it served for me as a really good general interest sort of background

and summary of AI technology and how we've gotten to this point. And believe me, I know there's certain of you listening right now in the audience that this is what you do every day. You're working at deep mind or whatever. But for the rest of us, one of the things that I liked that you did for me, you know, we hear buzzwords like machine learning and deep learning. And there's all sort of, I mean, we can get into the weeds about how these things are different. But in essence, the holy grail that everyone in AI is working toward is general artificial intelligence, right? AGI. What is that exactly?

Yeah, and first of all, thank you so much. I so appreciate that. per your question, I think part of why I wrote the book was that I am passionate about more of us being involved in these discussions and having a cross-pollination of ideas and an interdisciplinary approach to talking about these things because so many people say, oh, this is over my head. Or I already know this. And the truth is we should need to, we'll never get all the way there, but we need to start having more conversations across silos about how we can build these tools. So yes, in terms of what you said about that potential holy grail of AI, artificial general intelligence, you know, the idea of having AI that is smarter than humans in all ways and is smarter than we could ever be in all ways. Because we're already, per your timeline question, we're already there in terms of narrow forms of AI in our daily lives, from algorithms that Netflix and Google use, we're already there. So this idea of general AI that could potentially be smarter than we are or could ever be is both a technological question but also a really philosophical existential question. Or even, I'm sorry to interrupt, or even we're already there where obviously, you know,

AI can defeat us in games like Go and chess, but those are very narrowly defined where there's parameters, there's rules. And so that's still not the Holy Grail because it's almost like the AI has to be smart where there are no rules and parameters and boundaries and things like that. Exactly. And one of the things it's very difficult in getting us to that road. And again, we have no idea. if we could ever get there, let alone in terms of a timeline.

So that varies very widely across experts. Some people say it's definitely going to happen sooner rather than later. Other people say that could never happen. And one of the things, of course, in the way of that, is integrating into AI what we as humans call common sense. And so part of what my book is about is this idea that, you know, the human brain is this extraordinary mechanism, and we are trying to model AI on the human brain. But, of course, we know so little about the human brain in so many ways, why we sleep, why we dream, how we process memories, exactly how we make decisions.

So I kind of make this rather radical, and I hope it won't always be radical, you know, suggestion that our failure to model AI on the human brain might actually be our salvation because it will force us to take a wider lens and a broader view of what intelligence is or could be and to widen our lens and to give more dignity and respect to. all types of intelligence even beyond human. So for me, it's part of a bigger question, whereas, yes, that is seen as the AI Holy Grail, but maybe the Holy Grail is something even bigger than that and bigger, you know, bigger than any one tool we could build and more about who we are, why we're here, and that we share our very fragile, dying planet with so many types of intelligence. You know, I do want to come back to that in a second, but that makes me think, Let's go through the buzzwords real quick. When we hear like neural networks, that is, that's essentially the buzzword for we're trying

to mimic the human brain, right? Right. For the most part, though, technically that doesn't have to be the case. It's more about mimicking the vertebrate brain cortex. But yes, in a lot of the definitions, you'll see mimicking the human brain, which is such an important point, Brian, that you bring up, which is that we don't even have a universal definition of what intelligence is. And we use words like artificial intelligence, and I believe that term will probably eventually go by the wayside. Some people think it should be called

augmented intelligence. So these terms are shifting. I think it's part of why people think I could never understand this, but even within the fields, in terms of universal definitions, they can be elusive. So neural networks is a great example. Most definitions say human brain, but technically it doesn't have to be so. Well, and then I'm sticking with this theme of you explaining buzzwords to me. When you design a computer to be creative, I got this from your book, that's essentially machine learning in the sense that a machine doesn't have to follow rules. It can take in data, learn from that data, and create its own rules, right? Exactly. That kind of study of the things we call

algorithms and those models to perform those tasks. As you said, without those specific instructions. So technically something that could learn on its own without a human guiding every step of the process. And then deep learning is a subset or maybe one part or one flavor of machine learning. And then machine learning is one part of AI, basically. Yeah, exactly.

Deep learning part of that kind of family of machine learning. And exactly, as you said, and broadening that to that umbrella of AI. But again, things are certainly shifting in the space. But yes. No, but, I mean, again, that's what I found so valuable. Like, even I that have to talk about this every day, like, that was useful for me to get, like, the context of what the buzzwords mean. And then even the things, like, when I do segments and stories about how we're already at the point where we can't always explain how AI reaches conclusions on certain things and how AI does what it does, even.

the people that design the AI don't know. I did a segment recently about how, I can't remember, some AI can determine if you're going to die in the next year of like a heart attack or something, and the doctors don't even know how they do it. But that's the function of that creativity, of that machine learning, right? I think like when I speak to technology audiences, that is one thing. I'm constantly getting people saying, yes, you know, even as an AI developer, sometimes an AI developer can't understand how the actual AI it built makes those decisions, which is one of those really important touch points to understand that even the AI developers creating the tools don't necessarily know how they're coming up with decisions.

And this idea, of course, is you tapped into creativity comes up a lot in my book. I have a whole chapter about it because of the nature of human creativity in many ways is elusive. So we're trying to model this thing based on these huge. human attributes, and for me, we also need to be asking these questions of ourselves, because I believe AI and technology in general is a mirror for who we are. And so the question of could an AI be creative is also one that is very divisive, and people have very different opinions on if that's possible. But I do think that, you know, there's a lot out there about, well, we have to stick to human creativity because that's something a machine could never do. And maybe that's true,

but maybe that's not true. And then what does that mean about who we are and what our place in the universe is. And these are important questions we need to be asking ourselves. Well, and that perfectly brings us back to this idea of maybe you can model AI on intelligence other than the human brain. And not only, okay, based on evolution, the human brain is, for our purposes, at least, the best that evolution has been able to achieve in terms of intelligence. But there's no, that doesn't mean that going forward for the types of tools, types of AIs we want to develop that that means that the human brain should be the model. And what I loved, I love the chapter going into, you know, maybe you model it on insect and hive

brains or octopuses, octopi and cephalopods and things like that. It's a really important part of my book, a really unique part of the book. But again, I'm certainly not the only one thinking about these things. I would even go back and just say that, you know, one of the assumptions based on what you just said, that it's our best type of intelligence evolutionarily speaking. And I don't think that that's true. I think that, and again, going through the incredible intelligence of Billy the octopus in the Seattle Aquarium or insects or ants or termites can do extraordinary things the human

brain simply cannot do. So one of the big premises of my book is that I suggest we widen our lens to say, well, the human brain can do extraordinary things, but an octopus can do something we can never do. And for example, scientists at Raytheon are looking at the octopus distributed approach to intelligence as actually something better suited for the robotics they're building to explore distant planets. Or Rodney Brooks, who has been inspired by insects in his robotics work, or Dr. Radica Dirks is doing interesting work with schools of fish. So there are examples that abound. And for me, the question is, what happens when we take away that word best and open ourselves up to say there's all

these different types of intelligence, all unique and all special, but for me, one of kind of the baseline core ideas of the book that comes a lot from my experience as a human rights lawyer is that exceptionalism is dangerous, and you don't have to be exceptional to be unique and rare and special, and you don't have to be the same to be worthy of value and of dignity. Well, and also, I think a key thing there is also best, but also best for what we want AI to do. So again, right, there's a certain egotism to that, like, well, we'll just model it off of ourselves. But modeling off of ourselves, you know, the human hand is great, but it's not the best tool for doing, manipulating certain things either. And so also, I mean, that's a huge part of your

book, too, is how you talk about the need to take an interdisciplinary approach to building AI systems. And we've actually spoken to people about that on the show before. But so speak more on that about how we need to broaden as opposed to narrow. Don't be narrow. Don't use previous examples. Try to look for everything in terms of finding a model for the future. Yeah. And again, it's only natural. I mean, technically, as human beings, we could only really ever come from our own experience, right? So, of course, it's not so not even that it's difficult. it is impossible to take the point of view of another species that we are not. But the point is to strive to try to kind of widen our lens is always been beneficial

throughout time and space and history being exclusive and excluding groups. And of course, that includes groups of people, has always been limiting at best and incredibly dangerous at worst. So looking at kind of the vast panoply of life and intelligence we have on this planet It teaches us so much about ourselves, but also the world around us, how to protect it, and how we might model kind of the next generation of intelligent technology. And we have so much to learn when we're willing to say, as you said, well, we might be the best at this, but there's something else that might be better for these other purposes.

And inevitably, creativity, human ingenuity, and our collective intelligence comes from, one of the ideas I bring up in the book, this idea of combinatorial creativity, that people like Leonardo da Vinci had, that you're taking disparate elements from different disciplines and combining them into something new. Because AI in and of itself is a discipline that draws on psychology and biomimicry and neuroscience and computer science. So, of course, also in the book, as you saw, one of the things I suggest is that we need to be working across silos, and our educational system needs to include a lot more elements working together. And that's even just to crack the code of AI. One of the things that we need to be working,

We've ended up talking about a lot recently on this show is the sort of either Cold War or dichotomy between China and the West in terms of technology broadly and in terms of ethics and morality broadly, but also about AI specifically. like, you know, we've talked about how China sort of had this AI Sputnik moment when AI defeated the greatest human go player in the world. And so I think I read this piece on the show. But like, there was a piece that I read recently from, I believe it was a Chinese or quoting a Chinese AI researcher where, like, he pointed out that, for example, the Chinese AI researchers are willing to accept a bunch of more accidents and even more deaths at the beginning of, say, self-driving car research than we would in the West, where we don't even – we can't accept even one. And so, again, as with your background in human rights and things like that, I'm curious

your take on China and their AI developments and how ethical, and even morally, it's different than maybe what we would hope for. Indeed, yeah. And this is a very important question, and I have spent most of my life living abroad, and I've spent a lot of time in China, and I used to live in Hong Kong, and I teach a class on ethical decisions and leadership in undergraduate class, and also at the graduate level, and we discussed this a lot. And, of course, it has to do with the famous trolley car problem of, you know,

if you were able to determine if the trolley car is diverted to another track and kills one person versus staying on the track and killing 10 people, what would you do? And there's endless permutations, but it has to do with ethics and morals and values because, of course, one of the questions is, even if we could build ethical machines, whose ethics would we choose, whose values, whose morals?

And these are some of the oldest questions we've asked ourselves in philosophy and in human rights. And every culture is going to take a different approach, which is why collaboration transnationally across borders, is so essential because the short answer is we'll probably never get all the way there in terms of picking one universal idea we could all agree on. Though, as you know from the book, I do propose the international human rights framework as something that's been probably as close as we've gotten as a human society of figuring out one ethical code. But as you said, China's going to do

things very differently. Everyone in every culture is going to do things differently. But I think that it starts with asking ourselves the question, because as of now, we're having little debate about even asking that question. And I think that there is no one answer, which is why the trolley car problem and should we kill one person to save a million, would you torture one person to save a million? These are questions the US faces and every country faces all the time. There is no one answer, though it starts with having a discussion that the ethics matter when we're thinking about things like self-driving cars.

And also it helps bring more people into the room. example that a lot of people might be familiar with is this idea of the faucets, the automatic faucets you put your hand under, you know, that are now everywhere. And when they first piloted the touchless faucets, they were only primed to white, pale skin. And that wasn't any malicious reason, but it was because there was no person of color in the room to be able to say this doesn't work for me. So we don't have answers to pretty much most of these questions, but one thing we can

do is work on more diversity, more inclusion, more representation, more decolonization of how we build these tools. And in terms of China, of course, there's a lot about the arms race, and that is really real and really happening. And China's ahead in a lot of these ways because of what they're doing and how they're able to manage it. But there are examples of China and the U.S. working together, and I'd like us to see a lot more of that. The book, again, is a human algorithm, how artificial intelligence is redefining who we are. Flynn Coleman, again, this is, in terms of the books that I've read over the last few years about AI, I think this is, if you want to get the best grounding and understanding of where we're at,

I think this is the best book to do it. I'm so honored. Thank you so much. That means a world to me. It was a delight to be on your show.