Into the Impossible With Brian Keating - How Will AI Change Science? | Brian Keating & Cassandra Vieten (#379)

Episode Date: December 19, 2023

Remastered from my discussion with Cassandra Vieten at the AI & Your Life - The Essential Summit. Will AI bring us closer to the truth or further away from it? Is it changing the way we do science? And will it eventually replace teachers and professors? I had the pleasure of discussing this with Dr. Cassandra Vieten at the AI & Your Life - The Essential Summit.  The AI & Your Life - The Essential Summit aimed to provide a comprehensive guide to artificial intelligence that avoids the hype and gives you the practical knowledge you can apply to your everyday life.  Key Takeaways:  Intro (00:00) AI’s role in redefining the epistemological foundations of science (00:44) Introducing unconventional research methods (09:49) Will AI replace human educators? (16:31) Can AI be conscious? (20:20) Will AI bring us closer or further away from the truth? (30:00) AI and spiritual insights (39:35) Outro (42:40) — Additional resources:  📢 Ownership of your health starts with AG1. Try AG1 and get a FREE 1-year supply of Vitamin D3K2 and 5 FREE AG1 Travel Packs with your first purchase 👉 https://drinkag1.com/impossible ➡️ Follow me on your favorite platforms: ✖️ Twitter: https://twitter.com/DrBrianKeating  🔔 YouTube: https://www.youtube.com/DrBrianKeating?sub_confirmation=1  📝 Join my mailing list: https://briankeating.com/mailing_list  ✍️ Check out my blog: https://briankeating.com/blog.php  🎙️ Follow my podcast: https://briankeating.com/podcast  — Into the Impossible with Brian Keating is a podcast dedicated to all those who want to explore the universe within and beyond the known. Make sure to follow so you never miss an episode! Learn more about your ad choices. Visit megaphone.fm/adchoices

Can a computer replicate a visual system? Yes. Can it replicate memory? Yes. Can it replicate processing power? Of course. It's clear computers are better at us at chess. Will a computer ever create a game like chess? It will get to be an approximation. And the question is, is the approximation at some level equal to the human experience of consciousness? Any sufficiently advanced technology is indistinguishable from magic.

Open the pod bay doors, how? it's really great to be here with you. It's always great to be with you. And it's just lucky to have you as a colleague and friend here in San Diego. Yeah, yeah. It's been great. So, hey, everybody's talking about artificial intelligence. And some people are talking about it with great excitement.

And some people are talking about it with a lot of fear and everywhere in between. How do you see the role of artificial intelligence, especially in shaping or even redefining the epistemological foundations? of science, like let's say in fields like astronomy where data is huge and abundant, but often needs sophisticated interpretation. It's so hard for me to decouple the just outright joy and fun and pleasure that I'm getting from AI explorations and experiments, you know, from the legitimate concerns that I think people need to be cognizant about.

It's kind of a threat and an incredible, an incredible. toy and opportunity for so many of us, especially those of us with either children or those of us with educational children that we get to teach. I really loved it. It's been kind of almost like getting to redo the internet boom of the 90s for those of us that are old enough to remember it, but didn't really participate in kind of the shaping of it. I was too maybe young at the time. But nevertheless, this is kind of a new opportunity when we have an opportunity to shape the future of education, of research, of scientific, searching for answers and truth, but also redefining what it means for us to be human beings. And fundamentally, I think people lose track of the fact that because of my colleagues are so otherworldly brilliant that they actually are not human beings. But the truth of the matter is science is done by scientists and scientists are human.

And we have all sorts of wonderful qualities. We are inquisitive. We are imaginative or curious. We love to, you know, kind of research and play around. And we're also like children in another way. We don't like to share our toys. We love to claim credit and get attention. And we like to have the shiniest new thing on the block.

And so AI is all of this at once. It's a tool. It's a helper. It can be sort of dominant in what we're doing. And my question is, how do we get it to serve us? You hear a lot that AI isn't going to replace your job. You know, somebody who knows how to use AI is going to replace your job. I actually don't think that's true, especially in the professorate, which is my occupation.

I actually think we're completely immune from AI. And that's a shame. I think that's a true failure of academic. more than anything else to adapt and adjust and recognize how challenging the future may be when an AFAI can be brought online in a well-aligned way with a teleological purpose to improve the educational outcomes for our students and make education more affordable, more accessible, more democratized for a greater number of people, because I believe the only way to really have any hope for humanity's distant future

any way that we could last even a tenth as long as we've been on this planet, which is only, you know, quarter million years, the only way we can get to, you know, the fraction of that is by really pacing in a proper way our technological understanding of the cosmos. So I'm excited about it. I think it's one of the most kind of energizing aspects of, you know, technology that's come along in a long time. I think we've only scratched the surface of it.

And I'm excited to really apply it to really, at some ways, you know, make my current job obsolete. And maybe we'll get into ways I think we can do that. Yeah, that's great. I feel a lot of the same ways you do. My dad is a scientist. And when I was growing up, we were really early adopters. And, you know, we had the gigantic BCR with only one movie, you know, at home. And the, you know, Apple 2.

I think we had the Commodore 64. You know, we had the Apple. hasn't. Yeah, right? So I mean, like, we were adopting everything early on. So when I first heard about AI, I was like, this is awesome, you know? And at this point, I keep it open on my browser pretty much all day. And I don't use it really to, at least the chat GPT form of AI, I should say, I don't really use it to copy and paste from, but I definitely use it if I get stuck, you know, writing something or thinking about something. And I found it just a great springboard and almost like a partner. You know, it's like data in Star Trek, you know, I wouldn't take a data if I could have one,

absolutely, 100%. So that's great. Let's go to the positive side first. I mean, the positive side, like you were, you know, you work in astronomy, which is massive, massive, massive data sets. We also know that medicine, for example, there's a lot of promise now of AI being able to assist in detecting disease very early from these massive amount of, amounts of data that it's just very, very impossible for humans to be able to synthesize that

much data quickly. Yeah, I think having access to a device that has the entire corpus of human knowledge and instantaneous recall and just incredible synoptic capability to cross-reference all the knowledge that's ever been generated, I think it'll be almost a form of malpractice for physicians not to use it for pilots, not to, I fly a little planes around Southern California. And you'd be terrified to know that not only in my little Cessna, but in the 737 or whatever you fly in to go to a conference on Southwest Airlines, you would not believe the technology that's in the cockpit is from the 1960s. Some of it's earlier, 1940s, the fundamental radio transmissions that guide and communicate,

they have not changed at all. And so what happens in the cockpit? And I'll only speak to this. I'm not an expert and, you know, I'm not a physician, even though I'm a doctor, you know, I'm not, I would say I'm not the kind of doctor that helps people, maybe helps people learn about science. But there's a pilot sitting in the front. And, you know, when they, before they take off, they're getting this information from a control tower and only one person can talk or listen at a given time. It's a one way, one, you know, channel communication. The biggest airports in the world doesn't matter. One person talking to one other person. If there's an emergency in another plane, no one will know about it. It's terrifying. Then you're getting to the

airport and there could be like a, you know, a plane that popped the tire on the runway. You won't know about that for about an hour because it takes that long to report back. There's no live monitoring. And then furthermore, you have to sit in the cockpit and twiddle a little rotary knob, dial in a discrete frequency on the FM radio dial that only pilots can access in their aviation cockpits. And then when they get to the station, then you have to wait a minute to hear the loop of the recording and what's going on at the airport and all this and you have to take your eyes off the outside world as you're flying and the controls. I'm sure if people are scared of flying, they're going to be more scared now. But it's 100% right for automation. As is,

having a, you know, a data, you know, like, you know, the movie hurts, great movie, you know, in your earpiece, the doctor should have that because the patient will be talking and the patient's looking up at the ceiling and the doctor's looking at, you know, her computer typing in a bunch of notes. They're not looking at each other, and there's tons of nonverbal cues that are being missed that a camera might pick up or, you know, from subtle things that no human can pick at the gate, how they walk in, you know, how are their, you know, pupils dilated when they mention things, has their skin temperature, all this stuff, and it's just going to waste, all this

information. And not to mention, you know, whenever I, you know, sit up and bed, I get this pain in my left pinky toe. Again, I know nothing about medicines. Okay, so you sit up in bed, you get a, oh, that's, you know, synoptions disease. And, you know, that's fatal 100% of time if they're not diagnosed. And the doctors, oh, don't step on that toe. Don't do a sit-up anymore.

But, you know, computer can then synthesize all this knowledge, provide a synopsis of it, deliver it. Now, there is an impediment to all of this in aviation, especially, and I imagine in medicine as well. And they're entities that are, you know, that are humans, they're called lawyers. And a lot of this progress in aviation is stymied by the fact that if there's a lawyer involved, there's an accident involved. We live in an extremely litigious society. So innovation is stifled. And despite the fact that there be a net tremendous savings of lives in aviation and just overall safety culture and near misses counting as accidents,

you know, just avoiding those altogether in the medical world where it's prevalent as well. And we're just not doing it. And I can't help but think it's because of something very. very, very non-technical called lawyers. I mean, you bring up a really good point, which is that there is this incredible ability to synthesize, but there is the risk that it ends us up with the lowest common denominator, right? So you're someone who values both consensus wisdom and non-consensus insight.

How do you reconcile the potential of AI to either enforce existing biases in scientific research or on the other again, introduce non-conventional methods that could be game-changing. Yeah, I mean, that's one of the things that I've rarely been playing around with early on in the AI kind of revolution, which is only like eight months old. I mean, I was playing around a little bit with it. And I've always been interested in it. I've talked many times with Sir Roger Penrose, who wrote The Emperor's New Mind, which talked about computers and minds and brains. That was the first popular science book I ever read in 1989. when I was in high school, didn't understand it then, still struggle with it today.

But talking with him about what is a brain, what is a computer, what is intelligence, what is imagination has been just phenomenal for me. And I've always been interested in that, even though I study, you know, astrophysics in the origin of the universe. I think ultimately they are related at some level. But one of the things I play around with the most are these, you know, scientific co-pilots and tools that allow me to read and to extract information. And I don't use them when I construct lessons for the primary lesson.

What I've been doing is despite the fact that UCSD is the best students in the world, some of them come in, even in advanced cosmology, and they'll have some lacunae in their understanding of some topic in electromagnetism or special relativity, just something I might have covered three years ago. Maybe it's not so fresh. Everybody needs it. So what I'll do is there are these tools, and I'll use AI, just generate a whole bunch of slides with the bare, you know, kind of minimum of partial differential

equations of special relativity of Maxwell's equations and electricity and magnet. Just synthesize those into a slide deck. No, it's not mandatory. It's not part of the main class. But now I can have access to these students. They can have supplemental conversations with these documents. You can now have a chat bot that basically for each of one of my Keating's supplemental, you know, strategy lectures, they can then chat with it and they can interact with it. And that's, That's been a lot of fun. Another thing I've done to synthesize, I would say a synthesized consensus, but to understand how it's possible that you can at least come up to speed with maybe an adversarial point

of view. Science is very adversarial. The scientific method is one of natural, hopefully good-natured antagonism, but nevertheless, you should, you know, especially in fields where human lives are at stake, it's something that's very precious and you need to be careful with. So what I'll do is I will take a database of writing from great old works, which I think are still pertinent and relevant today. And I've started to try to make interactive chatbots with the entire book of Galileo Galilei,

who's my intellectual and scientific hero and mentors me from beyond the grave. And the way that we do this is I have access through the University of California, which allowed me to use the rights to Galileo's dialogue, the translation of it. And I made the first ever audio book of that book with my friends Carlo Rovelli and Lucio Picciarillo. And it's a play. I mean, you're basically three characters over three days interacting with, well, trying to figure out is the earth, the center of the universe? And now we just take it for granted, although I'm asking you, I'm looking at you, not you, Cassie, but I'm looking at the audience right now. I'm saying, could you prove standing on one leg that the earth goes around the sun?

It doesn't look like it to me. Looks like that bright yellow thing is moving around us. So challenge yourself. Don't just accept this accepted wisdom. And that's another thing that I use these, you know, talk to me like I'm five and prove, you know, that the earth goes around the sun. These are things that, you know, people just take for granted. So I took the works of Galileo, digitized them, they were digitized, and incorporated them into a large language model. So we have the first, you know, chance to chat with, I call him Galileo, so artificial intelligence, Galileo, and interacting with him over the centuries.

And you're having this conversation, as Carl Sagan said, a book is proof that humans can work magic because you're having a conversation with a long dead author. That was a book. You know, he never lived to see podcasts like mine where I've had his wife and his daughter on, but he's never, you know, he never lived to see an interactive chatbot. And so now we can have Carl Sagan bot and we can talk to him and ask him questions, not just scientific things, but also use him as your research assistant, but simultaneously as your research mentor. And these are things that I think we need to as educators incorporate and inculcate before it's too late because the academic model goes back even farther.

The one that we are using today at UCSD that you and I use with our students is a thousand years old. It's 500 years older than Galileo. It goes back to the University of Bologna, Italy, and the year 1080, almost nothing has changed. It's a sage on a stage and you and I taking a piece of rock and scraping on another big piece of black rock and the students are falling, you know, and they're, you know, they're not getting any update. Galileo could walk into our classroom right now and he would recognize exactly what we're doing. It wouldn't be like, you know, there's some mystical orb glowing and transmitting knowledge holographic. No, no, no, it's just a person scraping on a rock.

That's all we've come up with in 500 years since he was born. It's crazy. So I really do think that it's ripe for disruption. I'm hoping to be a part of it, but it's exciting and energizing at the same time. Hey there, fellow Voyagers into the impossible Tis I, your fearful host. Professor Brian Keating here with a tiny little homework assignment before we get back to the episode. And that's to make sure that you're. subscribe to the podcast, either following it or subscribing to it depending on your podcast

catcher of choice. I did some research of my own and found out that only about half of you are actually following or subscribing to the podcast. So please do that. And for some extra credit, if you're looking to boost your position on the grading curve, please leave a rating or review. It really helps us out tremendously. Do it. Do it now. Before you forget, let's go back to the episode. And then when you think about teaching, I mean, do you think AI can ever replicate the sort of nuanced approach that a human educator brings to teaching these complex subjects? It's hit or miss. It's always augmented. Yeah.

Yeah, I find a very hit or miss. Not just the hallucination problem, which is truly a problem. You know, I like the hallucination. It's sometimes because I'll ask, you know, tell me about Brian Keating. And I'll say he wrote, losing the Nobel Prize. And he wrote into the impossible, and he wrote a brief history of time. And I'm like, no, one of those things is not true, but then you need to figure out, now you need an adversarial chatbot to check on that chatbot.

So there are things that are challenging and obviously unacceptable for use and kind of these objects are larger than they appear in the screen as a threat to blind application of it. But on the other hand, it does allow for the processing and sort of the synthesis of a tremendous amount of historical knowledge where it's bad. And I think that I am an AI optimist in that I don't worry about things like the so-called paperclip problem that you're just going to optimize and maximize. My friend Michael Shermer always says, you know, you drive a Tesla. It's got autopilot, which is an AI-enabled interface.

and driven by not a language model but by you know artificial uh intelligent engine in machine learning and machine vision etc and it knows you know i'm driving he's driving from santa monica to santa barbara and the quickest route is actually to drive across the you know to drive across malibu bay and you know drive on the sidewalk and mow down some pedestrians but it doesn't do that um so you know optimization has to be sort of shepherded and i always point out the the one statement to me. I know you're writing a book. I can't wait to read it about imagination. Now, Einstein said a lot of things. One of the things he said is imagination is more important than knowledge. I always say, you know, like, that's great, but I don't want my, you know, my surgeon to say, oh, I'm going to try this creative

new procedure when I give you the Brazilian buttlift or whatever I'm going in for, right? You wanted to like use knowledge, not wisdom or not imagination. On the other hand, he also said that his happiest thought was when he envisioned himself freely falling in an elevator or out of a building. And in so doing, he would experience no gravitational force. I always point out the chances or the likelihood of a computer, A, understanding happiness and like why that was kind of giving him a sort of pleasure and how that pleasure was connected viscerally to the preconception of falling that led to this creative, adjective breakthrough is

almost impossible. You can supervise, you know, making, not hallucinating, but you can't supervise that creative. You said this place was steps from the water. We just haven't found the steps yet. How much did we save? Enough.

Enough to get lost. Or you could book a stay with Hilton. Welcome to your oceanfront room. Just steps from the water. The Hilton sale is on now. Book on Hilton.com or Hilton app and save up to 20% to get the stay you expected. When you want savings, not surprises.

It matters where you stay. Hilton for the stay. Emotional thought. So I've been thinking a lot about that. I feel like it's not as much of a threat as sort of the doomsayers, Elon Musk's and so forth are stating it to be. And I think that they look at things as an engineer and they really don't look at things from a physicist point of view. We have to look at the exchange of energy and balance of resources. We have to look at physical limitations.

Those are the only laws of nature that are inviolable. Instead, they kind of look at it as, you know, like there's going to be some kid in a hoodie in Palo Alto and she's going to be, you know, just programming away and turn us all into paperclips. So it's very different when you look at things from a physical limitations perspective versus an engineering possibilities perspective. And I think we need both. Yeah, yeah. I mean, I think that big fear, right, are the biggest fear,

is that AI's would, you know, AI would come sentient. You know, you and I first met when we talked to, when we were working on, we were talking about consciousness together. And so, you know, if you look at biology and humanity and reality from a very mechanistic perspective, then you think, then the brain is just a very complex computer. And we're worried that this other, even more better complex computer, will somehow enslave us or will somehow,

how start talking to each other and decide, you know, we've been given the task of making the earth a better place and the obvious number one way to make the earth a better place is to eliminate humanity. That kind of... Are they brains? Are brains computers? Are they... Is AI conscious?

Could it ever become conscious? There are open questions. I think first of all, you have to recognize that there is no real understanding of consciousness among conscious entities like us. There are people that suggest that everything is conscious. There have talked to people like that, and there are people that suggest that, you know, evolution has produced the sensation of consciousness, but it's not necessarily real. The so-called hard problem of consciousness is coined by one of my past guest, David Chalmers,

and basically, you know, suggests that these quality of the experience of another individual is impossible to understand, to replicate. And these things go back way before David Chalmers. I mean, their famous essay in 1971 by Thomas Names. was entitled, what is it like to be a bat? And the answer is, I don't know. And neither do you. And this, you know, 52 years ago. So the question of what is consciousness, or rather, the question of can a computer be conscious is impossible in my mind to define before we understand what consciousness is. So I just stick to easy things like, you know, special relativistic astrophysics. But I think there is a,

you know, there's a quality and a quantity to neurological processes to conscious processes. Certainly, you could have like this incredible experience of, you know, kind of equanimity and so forth and be meditating. But then if you have no memory, if you literally have no short or long-term memory, what does that mean? So is that entity conscious? And then if you only have a memory, but you really don't process the state of awareness of stimuli. Or conversely, if you're overwhelmed by the stimuli, like if you have ever done this, you close your eyes, you know, you think, oh, I want to close my eyes, everything's going to be dark.

So everybody out there, close your eyes. And then notice, as you're closing your eyes, that there's a tremendous amount of light-like sensations that you're actually perceiving. And yet your visual system is completely shut off. And you can even do this in a completely dark room. You could go in outer space and this will happen. What is that? What's going on there?

I mean, there's so many different processes. So there it's all awareness and you're aware of it. But, you know, again, you have no memory of it. Or if you only memorize it, just like you cannot, you have this persistence of vision in your acuity and your visual field and you cannot ignore things. You could also go kind of crazy. And then, you know, what level do you have consciousness there? So now, can a computer replicate a visual system?

Yes. Can it replicate memory? Yes. Can it replicate processing power? of course. I always say, you know, it's clear computers are better at us than at chess, but, you know, I can beat a computer half the time at Tick-Tac-Tow. But no, in all seriousness, I always say, I don't care that so much I accept that computers are better than humans at playing chess, but will a computer ever create a game like chess?

Right. Once you give it a set playground, a sandbox, if you will, then yes, it will exceed an Excel in certain capacities. You know, right now I can't cross the road. You know, I mean, when I drive in a friend's Tesla, you know, half the time, it's about to kill me. And I'm like, how do you, will you pay money for this little feature? I mean, it's $10,000. And it almost killed us.

And so, and that's like the most advanced system on it. Now, will it get better, of course. And you see these movies of Boston robotics and doing it all across the crazy stuff. Yes, it's getting better. It can sort colors. And it will get to be an approximation. And the question is, is the approximation at some level, equal to the human experience of consciousness.

Well, and can it ever get to the point where it either can be programmed to or it can somehow evolve to make voluntary decisions? Yeah, it's already doing that, right? So the reason that OpenAI did not have data before 2021 is because that was when their first AI forays were really starting to blossom and they were incredible and they were doing great things. So now there's this whole corpus of, you know, large. language products that have been processed supervised by humans somewhere at some point

with some boundary conditions. But now you've got AIs that are learning from what AIs have put out, but it's not clear that that's a convergent phenomenon that may diverge. I called it Mad Bot disease. You remember mad cow disease is like cows, eat other cows, and then they have these brain damage effectively become zombie cows. So when bots do that, I mean, they're already hallucinating even before they. they started to access the web before 2021.

So it's not all clear that just scaling up, this is awesome. I asked for a gluten-free recipe for 16 people on a Thursday night in the month of May where we wanna have whatever, and it gave it to me in literally a second. It's not all clear that that scales up completely once you start feeding in the large language model training.

And there may be kind of sandb-bought, or rather, there may be boundary fence is put around it to prevent the hallucination. That's a huge thing. Yeah. Yeah. And you actually said that you have both a high degree of skepticism and the belief that everything can be improved.

And so what kinds of, I don't know, levels of like scrutiny will we need to implement with AI in scientific research and education? So there are a couple of things that I think I would feel comfortable about applying AI, you know, even as it is now. And those are things like refereeing papers. It's an incredibly thankless job that these journals somehow only in the last 50 years. I mean, journals of peer review and we take it as the gold standard, not at all, you know, true or really present a couple of, you know, decades, you know, before we were born. So the fact that it's sort of like, oh, only humans can do this.

And that journals got people like me to do their work, you know, paint their fences for free, never been compensated. even, you know, referring nature articles and, you know, taking weeks of time to do it and very thoughtful, I think that LLMs can do a tremendous amount of good there because it's very proscribed, very clear to see what the boundary conditions are, how it can be used, and how it could be not to let it be abused, and do things like plagiarism searches, looking, searching for pe hacking, doing all the sorts of like gut checks that scientists don't really get to do because it's so hard to get the, as you know, to get the funding, write the proposals, get the funding with a 12% acceptance rate in my field,

maybe lower in other fields. Most people don't get their first research grant independently, NIH grant, until they're in their 40s. It's not like, oh, well, I'll have kids when I'm 40. That wouldn't have worked out so well. And so the question is, how do you actually make these tools be of service to you? I think Clearcut, and I read a study recently from Cornell, that they are actually, you know, at least 30% of, so I write a paper,

you know, but the Big Bang model and some new discovery that I may submit it. And then a person will referee it and then an AI will referee it. And then the author will get the paper back and say which one of these reports is more thorough, more helpful. And 30% already, like today this Cornell study showed, are better when the AI does it. Now, not 50% or more, but eventually it'll get there. And I think we can do away with this, you know, kind of, you know, basic the indentured servitude, you know, they guilt trip us into refereeing their

articles that they can put behind a paywall. So I think that's coming to an end. I think there'll be other ways to sort. I think we're going to get a lot of anything where a blind, you know, kind of audition, so to speak, applications to colleges, those kinds of things, submissions, you know, those can be reviewed and pre-processed and triage. That's very helpful. So, and these things are just fun. Like, it's really fun to play around. You know, my art My artistic skills, you know, are limited to like stick figures and they still... It's peak pollination season and my business is scaling fast. To keep the nectar flowing, I need a phone plan with top priority data speeds.

That's why I chose GoogleFi Wireless. My connections stay strong even when the hive is buzzing. Plus, unlimited plans started $35 a month. Now that's a deal that doesn't stay. Explore GoogleFi Wireless plans today. Plus taxes and government fees. Google FiWireless is not subject to data. traffic deprioritization during times of high network usage.

But nevertheless, I am a Da Vinci when it comes to using Leonardo or Dali or whatever. And that's been really fun for me, because, you know, especially making YouTube content. And I want to show, you know, like what Elon Musk claims he wants to, you know, die on Mars. So, you know, I'll make a YouTube thumbnail of a video where I'm talking with Joe Rogan about, you know, Elon Musk. and the thumbnail will be like an elderly Elon Musk, you know, shuffling on Mars with a space helmet with a metamusel biscuit in it. It's so much fun. It really is.

And I love to teach my kids how to use it because I think they can get on board really early. And, you know, it's so called prompt engineering, which is supposedly in the field. I love it. Yeah, or even an art. You know, I've heard it called prompt craft, you know. Yeah. Yeah, the art of prompt craft. I saw a demo the other day that was a new virtual reality.

application, which I'm sure a lot of people are racing to do this, where you're literally walking through essentially the holodeck and you're like, I want to be in a forest. And there it is. And I want the sky to be pink. And now I want a dragon flying through it. And now I want, and I mean, it's appearing as you're saying it. It's crazy. It's just wild. Well, what do you think about AI as it might impact the scientific method? You know, you've mentioned something about this, like that somehow AI could impact the scientific method to make it better or to impact it to make it further from the truth. Yeah, I think, you know, I think the problem is that we really don't have a really coherent

supervision strategy of how we're going to really supervise and make sure that they're aligned, not the alignment of like human flourishing that, you know, Musk or, you know, Sam Harris or somebody would talk about, you know, I'm not talking that existential at this point. But at the same time, I think, you know, we want to treat them and have some caution for how we actually interact with them, understanding their limitations. But I think the bottlenecks, again, are going to be, just like in the aviation and medical examples I gave you before, they're going to be other humans. They're going to be not the artificial intelligence, but the natural stupidity, you know, and the kind of pettiness and grievance, you know, kind of focus that we're going to have, that these things

are, you know, taking our jobs or these things are, you know, putting people in danger, et cetera, et cetera. So I do think that it's possible for us to go a little bit overboard and feel like, yeah, these things are a net, you know, just a net negative. But that being said, I think, you know, we do need to have some sorts of safeguards. I'm not sure exactly what those should be, but I think they can be deployed in low-risk environments like teaching, you know, as I've already mentioned, you know, I feel like there is very, it's very difficult for us to make a case that I am a better professor than Richard Feynman or then I'm better than Galileo. And these are, you know, people are Sagan or, you know, Jane Goodall or, you know, so we have

access to thousands, millions of words written by these individuals, the greatest educators, in history. And, you know, we have like a master class with Jane Goodall, but you, but to take a whole course and really interact with her and maybe use it to kind of, as they say, you know, flipping the classroom. I never really got into that. You know, I was more often getting flipped the bird, but getting into it now, I could see it. Yeah, you could really do a lot of the work and problem solving in the classroom with the professor. But you could do, and just ideation and and and creativity and imagination in the classroom. And then you could do the kind of grunt work,

oh, here's this partial differential equation and we need to solve it in these boundary conditions and et cetera. I think those could be done offline with a, again, with Galileo or Feynman teaching you how to do the, you know, solve that equation. They might not be so good if you talk to them. We set up a Feynman bot on my website, Brian Keating.com. You can do a chat with my avatar, bribebot. I've taken my books.

put them in there, you can ask it what I think about aliens or the Nobel Prize or Elon Musk, and it'll tell you, and it's pretty accurate, it's not 100% accurate. But to kind of do the heavy lifting and sort of the grunt, you know what it's like. You teach a, you know, if you're teaching a class and you have the choice between teaching spelling or teaching, you know, the works of Shakespeare, it's always more interesting to teach a more advanced subject. Yeah. This allows us to shortcut that, to get to the good stuff with the students who are thirsty

to get to that as well. Yeah, like you said, get a ton of prerebring. requisites out of the way. I mean, I love the idea of being able to someday do statistics and just be able to use regular language saying, you know, how is group one different than group two controlling for the following variables and have something come up in one second? Like, oh, yes, sounds amazing. I loved what you said earlier, too, about like the, in some ways, we're just racing toward the future. Like, the technological advancement is at light speed. And then in other ways, it's really not. I mean, I was using my air fryer the other day and I was like, this is the first time I kind of feel like I'm living in the future. You know, you just put something in it. You close the door. You open it and it's perfect. But that's not very, you know, that's not a big deal. We thought we were going to be flying on air, you know, flying cars and cruising around on meditating skateboards by now and, you know.

Yeah, we have, look, a tremendous amount of technology. One of my friends, Eric Weinstein, I've had it many times in the podcast. He said, you know, if you're were teleported, you know, from the 1950s to today and you just showed up in the living room, you would say, hmm, the TVs are flatter, but almost nothing else would be different, right? It would be essentially unchanged. Now, I mean, that, of course, if that's your whole domain, yeah. And I said, you know, if you go to this apple picking orchard here in Julian, California, you will notice that nothing has changed. There's still, you know, people and kids picking apples by hand.

Why haven't we had, you know, robot lasers, you know, zapping things? And then the process, no, I mean, some things will never, you know, really improve upon. There was an old skit on Saturday Night Live called Deep Thoughts with Jack Handy. Yeah, I remember that. Yeah. You do? I would think you're too young. But anyway, I love it.

And he had a book, whoever it was writing it. And he said, you know, it's amazing to think that the very first fly swatters were nothing more than a stick with a large thing at the end of it to kill fly. I'm like, still is like that, you know. Right. I love that. Paper clips were once just wound pieces of. of metal, you know, of course, they still are. So the pace is slow on one hand. Software is cheap, you know, Mark Andreessen, as always said, you know, software is eating the world. But if you look at

it, you know, the biggest companies in the world are not software companies. They're, you know, Tesla, Louis Vuitton, SpaceX, and they are Apple computer, you know, they're not really, and Open AI is a blip. Open AI is smaller than, you know, some cryptocurrency, you know, pervaders, right? So software is great. It's very easy to produce software. I'm not saying it's easy. You can produce a lot of it. And then the question is, you know, what sort of technical debt do you have that you say,

oh, I'm going to get to this later. I'm going to document this later. I'm going to patch this hallucination feature. Now the stuff's getting real. And I think, you know, people aren't trusting it to like diagnose, you know, tumors without supervision. But that can come, you know, things that can be outsource. to things that are better at scale and quantity can sometimes, I would say, you know, quantity in certain amounts equates to quality.

Yeah. So, you know, the fact that this thing can catch, you know, a 0.1% with a 0.1% error rate, you know, it's not 100% and we'll never get to that, but it's, you know, it can do 100,000 of those. So, yes, there'll be some and then those need to be double check. So it'll be slow. But again, I think it's the law industry, lawyer industrial complex that's causing it. A lot.

Yeah. I mean, you know, there is that fear that this will take our jobs. But then I read an article the other day that was like, well, what if our job, what if full time becomes three days a week? Now, that's a very thing about, yeah, meaning, right? Meaning crisis, you know, like how are we going to have, you know, what's it going to mean? We're not all going to sit around and write poetry, right? Right. And, you know, I do feel like there's just a question, you know, when the steam engine came around, people were put out of business. But, you know, in the. plow came around and the, you know, the tractor there were plow, you know, horses were put out of business. You know, so the ability to generate and to retool a workforce is, uh, is an interesting point. I do feel like, you know, thinking about things in terms of like how much leisure it will give us. I don't think that, I personally don't feel like that's the metric we want to optimize. I think, you know, people can talk about some basic input and basic income rather and so forth for the neediest of cases, but otherwise, I think people, you know, get a great deal of meaning out of doing their labor, their work, whatever it happens to be. And I think we can use it

in certain industries to improve life, but to think that it will happen in a vacuum that we're going to have these robots and we're simultaneously going to increase minimum wage, you know, incredibly, just thinking first order without really thinking about the consequences of downstream, you know, a couple of weeks later that's going to result in, you know, just mass layoffs where people make zero dollars an hour. So are they all going to become, you know, software engineers at DeepMind? I don't know. Right.

I think a lot of this will be addressed in some ways by competition. I hope that we don't have monopolistic like we did with browsers or computers and stuff. I hope that there'll be a lot. And it seems like there will be. I mean, I'm surprised at late, you know, it seems like Amazon got into it, given that they had these devices. I have one in the background, which I changed its name so I can say, Alexa. I changed it to Cassie. No, no, I didn't do it. It just changed it to computer so I can

have computer turn off the plug and they'll turn off the light behind me. So these competitions, I think, will be beneficial to the consumer. Love it. Well, my last question for you is, you know that I have a strong interest in the intersection between people's spiritual experiences and spirituality, religion, and health and well-being. And just like you were talking about having the chatbots of some of the scientific leaders, I'm curious sometimes about whether we can use AI to have spiritual insights and to experience things. I mean, certainly VR, I think, is underutilized so far in not only being able to maybe listen to a meditation in the morning, but being able to go meditate in a Zendo with a teacher who's amazing or a teacher who's died, for that matter.

Yeah. That kind of thing. had any thoughts about. Yeah. I worked a little bit here with Neil Smith over the summer who works in our facility called the Sun Cave. And he was working on an artificial Gandhi that then is a, spiritual guru, right, as a world leader and many a great thinker, philosopher, et cetera. And he was, you know, basically took an, you know, Gandhi didn't speak very much to, he distrusted video. And so there's not much of his voice actually recorded and even less of him on video. He was very deeply distrustful of it.

And so there's a couple of minutes, maybe 30 minutes or less. And Neil's gotten in his team of undergraduates, including my nephew, who worked with him over the summer. And they took, they digitized that. They digitized his likeness, his face. But they made it, you know, to avoid the uncanny valley effect, they made them kind of a little bit, a little bit cartoonish. But they actually imported it into Unreal Engine and just, you know, high quality video game. They map like his tunic with like beautiful textures. It's really a work of art and you should contact him or maybe for the summit next year.

We'll actually have a conversation. You know, you and I can interview Gandhi. Yes, let's do it. And it takes voice you dictate, transcribes to whisper, uploads the chat GPT. Gandhi's whole corpus of writings is available, answers the question. Then his voice is synthesized and his lips move accurately. Now imagine that, as you said, with not just dead gurus, but imagine, you know, I could talk to my late father.

Yeah. I would love that. You know, I'd love to have a conversation. Also with him, they're not so much digitized, but to have that opportunity or like, could it be possibly a source of comfort, you know, not to mention a business model, but it could be that people, you know, upload their avatars of their husbands. And instead of talking to, you know, strangers on Snapchat, they're talking to their husband and or their wife or their child, God forbid.

So I think these things have a great potential. for improvement. They also can be negative, as you know. But I'm an optimist. I'm not worried about getting turned into paper clips. There's not enough metal in a rocky planet of the size of the Earth to create infinite numbers of paperclips. And I'm more worried about, you know, human stupidity or natural stupidity than artificial intelligence around me. Yeah. Great. Okay. Well, hey, it's great to see you, Brian. It's great to have you with us. Thank you so much for taking the time. And where can people find you? My website at briankeating.com and then YouTube, Brian Keating, pretty big channel, podcast into The Impossible and Twitter and Instagram. You could find me there.

Sounds good. Great to see you. Thank you so much for all your awesome work and really glad to have you with us.