Latest / Cultural Wars | America Out Loud News / Can we literally sniff out and snuff out disease through AI

Transcript

Can we literally sniff out and snuff out disease through AI · Cultural Wars | America Out Loud News

• 0:08
  Really, we've learned anything from these past couple of years, my fellow Americans is that personal medical freedom and liberty are in crisis.
• 0:19
  Almeric Rot Law Pulse brings together the top experts in health care related fields to keep you up beaten again.
• 0:31
  You listen to America Albot Pulse Radio on the iHeart Radio Network and on the America Outlet Apps for Apple, Android, and Alexa.
• 0:37
  Don't miss a moment join us every Friday at 5 p.m.
• 0:40
  Eastern.
• 0:40
  I'm Dr.
• 0:41
  Randall Bach here to unpack myths, explore innovations and give you actual insights into the world of medicine and health.
• 0:46
  I'm honored today to have with us Bassant Dar, professor uh at NYU, uh, the Stern School of Hmm.
• 0:54
  Uh you'll have to tell me.
• 0:56
  Um, and he's a mathematician, logician, uh, expert in AI, and uh he has written a book which is not out uh quite yet.
• 1:05
  No, it is out.
• 1:05
  It is out, it has been in the last two weeks.
• 1:07
  Yep.
• 1:08
  Okay, thinking with machines.
• 1:09
  And uh it's a very uh interesting book.
• 1:13
  He uh combines uh the best of of all intelligences in all intelligences, his own uh combined with uh anything artificial, but his is actually rather genuine, and there's a lot of uh life story there and applique that um brings uh some things that might be a little bit esoteric down to the human level.
• 1:30
  So I applaud you for that.
• 1:32
  That's been a great read.
• 1:33
  I have not finished every single word, and I gather there's uh you know, the last chapter is kind of a power punch, and I hope to um be enlightened by that, as I have been by the rest of your book.
• 1:42
  So thank you so much for being on.
• 1:44
  Thank you, Randy.
• 1:45
  Great to be on the show.
• 1:47
  So uh let you know, thinking with machines, thinking without machines, you you you you deal with all of it.
• 1:53
  Um, and you kind of in order to kind of classify the artificial intelligences, uh, you need to understand kind of the natural ones and the way that the artificial, you know, appliques, um, you know, layer and filter um knowledge, and therefore you have to kind of look at us in the way we do the same.
• 2:12
  Uh, what has that trip been like for you?
• 2:14
  How did you get started in it?
• 2:15
  And what were the kind of um touch points that kept you interested in it?
• 2:18
  As clearly uh your intelligence, you know, which is vast on the topic, um, somewhat magnified and glorified the topic.
• 2:27
  You know, I'll start by saying that, you know, what a long, strange trip it's been.
• 2:30
  And it has been a long, strange trip.
• 2:32
  Um for me, it started in 1979 and in medicine, interestingly, and you're an MD, you'll appreciate that.
• 2:39
  Um, pure serendipity.
• 2:40
  I was, you know, 23 years old, wondering what to do with my life.
• 2:43
  I was in this PhD program, and uh student comes to me, a senior doctoral student, and says, Hey, um, there's this guy in the uh in the medical school who's been building this system called internist for the last 10 years.
• 2:55
  Um I want to have him offer a course on AI, just you know, let's show some strength and numbers and and come with me.
• 3:02
  So we rounded up two more students and went up to this medical lab, decision systems lab up at the top of the medical school at uh the University of Pittsburgh.
• 3:11
  And the professor Harry Popel was on the phone.
• 3:14
  And so we were waiting outside, and there was this big screen connected to a computer at Stanford, and those are days of time sharing.
• 3:20
  And it was a physician, snow white hair puffing a cigar, talking to the system via his assistant, because in those days no one could type.
• 3:28
  And they were discussing a case, a real case, and there was a QA going back and forth.
• 3:33
  You know, the computer was asking questions, he was responding, and uh I thought that was really fascinating.
• 3:38
  And at one point, the computer asked him a question, and he said, Why are you asking me this question?
• 3:44
  And internist said, because the evidence you've given me so far is consistent with the following hypotheses, and this question will have been discriminated between the top two.
• 3:52
  And I was just floored, you know, that just like wow, how the hell is a machine doing this?
• 3:57
  I thought, you know, I mean, I was an engineer by training, you know, I'd solved you know, complex differential equations numerically.
• 4:02
  That was my notion of computers, that they were calculating machines.
• 4:06
  And so this notion in 1979 of a physician talking to a computer about a medical case, just completely blew my mind.
• 4:14
  And that was my aha moment, and that's what really got me into AI.
• 4:18
  And I decided that's what I want to do with the rest of my life.
• 4:20
  And you know, that was the initial the paradigm of expert systems, or you extracted knowledge laboriously from experts, and you they then you represented it into the machine, and with the expectation that it would now think about new cases and in this case diagnose new cases that it did not hadn't seen before.
• 4:38
  And Intel this was remarkably accurate, you know, it rarely ever, you know, came up with the wrong diagnosis as long as you describe the symptoms correctly.
• 4:46
  Um so that was my initial foray into machine learning.
• 4:50
  And you know, I came to NYU in 1983.
• 4:52
  There were a bunch of database people, they wanted to hire an AI guy, and you know, one thing lectured the next, and then 1994, I brought machine learning to Wall Street, you know, this is a new paradigm, and you know, um created subsequent to that a machine learning-based hedge fund that has traded systematically ever since, you know, including today, it sort of keeps, you know, uh generating signals and trading.
• 5:17
  Um, and then more recently I've gotten into smell, like, you know, trying to digitize smell with the objective of predicting disease.
• 5:25
  You know, there's this woman who can smell Parkinson's 100% accuracy.
• 5:29
  Um, and you know, so there's there's clear evidence that you know, old faction can play a huge role in in medicine.
• 5:37
  And so I've gotten into that with some of my colleagues at the medical school here, uh, using like deep learning kind of methods.
• 5:44
  And uh and also I've been using sort of modern AI methods to um, you know, stimulate the thinking of one of my colleagues who's an expert in valuation, you know, valuing companies.
• 5:56
  And so we've built this bot called the Demotherin bot that actually thinks like him.
• 6:00
  So that's in a nutshell, sort of my long strange trip.
• 6:04
  It's been, you know, almost four years now in AI.
• 6:08
  And so um I wrote this book called Thinking with Machines, because to me, that is the future of humanity.
• 6:13
  It's inevitable.
• 6:14
  It's about thinking with machines.
• 6:15
  There's no opting out.
• 6:17
  And so we should start thinking about some of the pressing questions around AI and and humanity.
• 6:22
  And that's what the book is about.
• 6:25
  No, it's an incredible thing.
• 6:27
  Um I uh I just kind of pick up on the sensory uh part uh smell.
• 6:33
  You know, there's you know, I I've long had dogs.
• 6:36
  Right now we are uh canine less.
• 6:38
  So probably K8 or seven, uh we've decreased from K9.
• 6:43
  Um but you know, dogs meet and greet on a on a sensory level that's different from ours, and they have a sensibility, no pun intended, um, of you know who's coming in the door oftentimes, uh, you know, hearing footsteps and cadence and so forth, they they are measuring things that um you know we don't do.
• 7:03
  You know, there's a kind of if you've seen the movie Daredevil um uh Ben Apple like I think he's he's blind, he's a superhero, and you know, it's kind of a trope that the blind guy has all these X you know accentuated other uh aspects of sensing the world around him, and oftentimes I use the example, you know, of a blind man in a hotel room, you know, has to feel his way around in order to sense where everything is and might have a better concept of where the furniture is and understand when things have been moved, an ashtray or a chair, whatever, because he bumps into it and whatnot.
• 7:33
  So there are different ways of kind of perceiving the same world.
• 7:36
  And you know, we uh often take it tacitly that we understand everything around us because we have vision and so forth, we have color vision, but you know, they're obviously clearly parts of the electromagnetic um, you know, visual or quasi-visual spectrum that we don't perceive infrared and ultraviolet, all the way down to radio waves and all that kind of stuff.
• 7:53
  Um so there's you know, if there were you know something that had you know the sensibility of smell on top of you know, kind of a vocabulary uh uh of a dog that can perceive things in their you know, other animals that you know find each other in the midst and the mist of of the jungle and whatnot, you know, some little insect, you know, sits out a signal and he's finding, you know, or she is finding the the exact same opposite, you know, like member in order to mate and date and and meet and greet and propagate and whatnot, you know, it's kind of there's a um, you know, a lot of signal out there that uh we don't necessarily perceive.
• 8:28
  And so, you know, you can find kind of an augmented self in a way.
• 8:32
  Um, but you know, just to stick on the this the smell part, you know, the the molecules are are, you know, I mean, I, you know, if you could maybe, you know, get like a Pavlovian experiment in a sense to kind of get the dog sensibility of what the smells are.
• 8:47
  I mean, it's kind of an interesting prospect.
• 8:48
  How do you uh hope to like bridge that uh bridge of the nose, um, you know, in in terms of um, you know, monitoring these things.
• 8:57
  There's kind of like a post um, you know, reflective thing.
• 9:00
  I mean, how do you actually parse that?
• 9:03
  Yeah, you know, that uh that's a great question.
• 9:05
  By the way, on my podcast, you know, Brave New World, I've had four um sessions on smell, by the way.
• 9:12
  Um with uh a guy called Alex Wilsko, who is the founder and CEO of Osmo.ai, which is the first AI smell company.
• 9:20
  Um, you know, the you know, the the several ways to look at smell, right?
• 9:25
  One is, and this is what Osmo is doing, is to go directly from molecules to perception, right?
• 9:33
  That this kind of molecule will smell fruity, or this kind of molecule will smell pungent, or you know, oh, this will not cheesy or whatever, right?
• 9:40
  Woody.
• 9:41
  Um there is actually a lot of action there that that is there in the there is an association between molecular structure and smell, right?
• 9:50
  Uh as described in terms of language.
• 9:54
  I've taken a slightly different approach, actually a completely different approach, which is um.
• 10:00
  If if you think of like machine learning, uh, it's essentially, you know, uh, you know, you learn from data that's represented as pairs and XY X describes something, why is the target?
• 10:12
  So, for example, uh, this is a cup, right?
• 10:16
  The cup is the description of the object.
• 10:18
  So the object is, you know, however you describe a cup in terms of its parts and the label is it's a cup.
• 10:24
  Um vision has actually really benefited from that because we all agree on what we're seeing for the most part, right?
• 10:31
  And so there's this database called Image Net, which really catalyzed research in vision because there was like a large standard database that you could test algorithms against, right?
• 10:41
  You could develop algorithms, you could benchmark them.
• 10:43
  So it led to this rapid progress in vision, which is led to driverless cars and great stuff like that, right?
• 10:49
  So vision as a uh as a perceptive sense has you know, we've had we've made tremendous progress in it.
• 10:56
  Smell is tricky because how do you really describe smell?
• 10:59
  Like, how do you describe the intrinsic smell of a lemon or uh orange or cheese, right?
• 11:05
  You know, we use language, but language is a crude approximation to describe smell, right?
• 11:09
  We can say that an orange is closer to a lemon than it is to a pizza, but that still doesn't tell us intrinsically what it feels like to smell a lemon or to smell an orange or to smell coffee.
• 11:20
  So uh, you know, the the approach that we've adopted is to use biology to actually give us the ground truth associated with smell.
• 11:29
  So, what I mean by that is if we could have, in this case, let's say mice, you know, we we observe the old factory bulb of a mouse as it smells something, right?
• 11:38
  So we you know, the mouse smells something, we observe the old factory signature, which is basically a time series response of these glomeruli in the uh in the old factory bulb that light up.
• 11:48
  And that's the truth.
• 11:49
  That is the ground truth or the label associated with that molecule or the mixture of molecules, right?
• 11:55
  So the approach is we look at biology, we look at the response of the glomerulus, and then we use machine learning to say what's the mouth smelling, like and so we can now predict that oh, it's smelling acetone, or it's smelling you know, uh, an aldehyde or something like that.
• 12:12
  So we made some progress with that.
• 12:15
  And now the next step is can we actually use it to you know uh identify disease?
• 12:21
  Um, you know, because you know, presumably a disease will generate a different sort of time series response in the bulb than a normal uh uh like a control uh sample, right?
• 12:34
  So in a nutshell, what we're doing is using biology to actually give us the ground truth associated with smell, and then using machine learning that operates on that image that we observe, you know, from the biological organism uh as the truth, and then we train machine learning models to actually predict you know what the mouse is smelling.
• 12:57
  So that's sort of it in a nutshell, but that's very different from like taking a molecule and saying, Oh, that smells fruity or orangey.
• 13:04
  Right, right, right.
• 13:05
  It's kind of I'm kind of reminded by the of the you know, the cochlea, uh, the hearing, you know, little snail we have inside our um uh you know skull.
• 13:15
  Um, you know, the the the uh our hearing, we have a drum and it gets the vibrations, and then it's the little kind of hammer, um, basically a kind of like almost like a finger on top of a drum when you bang a drum, you can kind of feel it vibrating, and that gets transferred.
• 13:32
  Um, and then it gets played out in this little kind of oil-filled snail, which has a bunch of hairs inside of it.
• 13:39
  And basically it's it's a snail shape because it'd be hard to lay the whole thing out.
• 13:43
  And then the sound waves play out, and there's certain patterns we've seen kind of the Fourier kind of, you know, analysis of sound whenever overlay of waves, we can get, you know, not just the pitch, but we can get the tambour.
• 13:54
  So even two people singing the same note sound different to us because there's different splays the way the the sound wave comes out through the vocal cords or whatever, how sound gets transmitted transmitted in a room, reverberation.
• 14:07
  So that gets played as a sound wave along these tiny little hairs.
• 14:10
  And I'm just in my own imagination, you know, the certain you know, wavelength, and so the distant hairs are going to get the wavelengths that are further along.
• 14:17
  So everything has a pattern, and then those tiny little hairs, each one is attached to a neuron, and so it's kind of doing a um its own little, you know, basically part-by-part analysis in any given second, so we hear something.
• 14:29
  It's an incredible thing.
• 14:31
  Um, so I assume there's, you know, if you have could dissect the cochlea and lay it out, you get the same kind of thing.
• 14:37
  I mean, we see the same thing out.
• 14:38
  If you ever seen this thing that like people have a drum and they have candles all lined up, you know, in a in a line.
• 14:43
  So you make a noise and the candles all have different heights because the sound wave transmitted over it, you know, literally around the air.
• 14:51
  And that's kind of the the way we can see, you know, sound waves.
• 14:54
  But it's it's a fascinating topic to think that, you know, probably is some something not quite the same as the cochlea laid out, but this kind of like, you know, uh, you know, kind of like uh an analog version of a digital, you know, am I saying that right?
• 15:08
  Uh, an analog version of a digital concept, you know, it's happening in real time across many dimensions, but there are still still finite aspects that are being perceived in any given moment, like the breakdown of sound that we can hear just as a sound, but it's really just you know, certain of these cochlea that you know, neurons are just striking in different ways, like a guy playing you know, five different xylophone rhythms on a xylophone with you know thousands of keys, and you know, he can play them, you know, third for 50 or you know, 500 at a time differently.
• 15:37
  But it's kind of a fascinating topic that you know, these things can be broken down.
• 15:41
  Um, sorry for that riff, but um it's a fascinating concept.
• 15:46
  You know, I I have to say I'm very impressed by the um breadth and the um you know grasp that you have of of so many different topics, you know.
• 15:55
  I was fascinating in the book um uh of you know, kind of different topic, but day trading, you know, which I you know, being being uh, you know, pseudo-knowledgeable about so many things.
• 16:09
  I just kind of fobbed off when I heard about like day trading.
• 16:12
  I'm bah you know, these people just kind of pursuing things, you know, there's all this mix, you know, the stocks go up and down, blah, blah, blah.
• 16:18
  You know, so you win some, you lose some.
• 16:19
  How is that going to make money?
• 16:21
  But I I was fascinated by your you know assessment, which is really that you're kind of chasing the human element in in smaller, you know, segments of time.
• 16:31
  I don't know if you want to talk about your um foray, uh, you know, as you you mentioned, uh, into into day trading, uh, the concepts and whatnot, and how that that you know parsed out, how you were able to do this fine dissection a little bit along the lines of you know, parsing out smell or whatever.
• 16:47
  Yeah, so you know, so that was actually a very interesting phase of my life.
• 16:51
  I was I I just gotten into machine learning.
• 16:53
  That was like the early 90s, and um I sort of sort of I sort of the data wave coming uh from a long ways off, right?
• 17:03
  And um, you know, that data was becoming available, and I'd had some success in you know, um, with a project that uh AC Nielsen, you know, the media company, right?
• 17:15
  They had a they had a division in Long Island called Nielsen Household Services, and they gathered lots of data about consumption patterns.
• 17:21
  They had they were tracking 50,000 households, and every time they bought something, it was tracked into a database.
• 17:27
  And so there's this guy who came to my office one day, he says, Hey, I believe you you're doing something in machine learning, and that's about finding patterns and data, right?
• 17:34
  I'm like, yep.
• 17:35
  And he said, I've got tons of data.
• 17:37
  We sell information to people, you know, you know, do you think you'd find patterns in it?
• 17:41
  And I said, sure.
• 17:42
  And he gave me this data and came back two weeks later and said, What'd you find?
• 17:45
  And I said, like lots of things, but I don't know whether they make any sense.
• 17:48
  He said, like what?
• 17:49
  And I said, you know, all the women in the Northeast do a lot of shopping on Thursdays.
• 17:53
  And he says, Oh, yeah, that's coupon day, you know, uh, what else did you find?
• 17:57
  You know, and I was just like ecstatic that the machine had found this pattern because all I'd really told it to do was to find unusual consumption patterns.
• 18:05
  And and, you know, so I was working on this genetic rule learning algorithm at the time, and it came out a whole bunch of patterns that made sense to him, right?
• 18:12
  And that's what fascinated me.
• 18:14
  And the next year, someone introduced me to this guy, Kevin Parker, who was running technology at Morgan Stanley.
• 18:20
  He'd been a very successful trader who believed in technology and had lunch with him one day.
• 18:23
  But I talked about my ideas about machine learning and finding patterns and data.
• 18:27
  And he was just fascinated.
• 18:29
  And he took me back to the office, met three other guys, and made me a job offer, right?
• 18:33
  And I was like, what?
• 18:35
  And he said, Yeah, this is much bigger than you realize.
• 18:37
  This is kind of the future of finance and trading.
• 18:39
  So we're going to take a break with uh Vincent Dark coming back for our next segment.
• 18:44
  He's a professor of data science and business at the NYU Stern School of Business.
• 18:48
  And so this is America Out Loud Pulse, always a beat ahead.
• 18:52
  Struggling with your health, energy and constantly feeling hungry, want to save money on the rising cost of groceries and dramatically improve your health.
• 19:00
  Go to chemicalfree body.com forward slash out loud today, get nature's super multivitamin, doctor formulated green 85 juice formula, empower your immune system, cut your grocery bill, and save 20% on your first order.
• 19:14
  Tired of harsh chemical toothpaste that do more harm than good.
• 19:17
  Upgrade your oral health with smile.
• 19:19
  The doctor formulated toothpaste from the wellness company.
• 19:21
  SMALL uses powerful peptides, remineralizers, and probiotics to rebuild gum tissue, strengthen enamel, and keep harmful bacteria away.
• 19:29
  And also leaves your mouth feeling fresh with a clean mint flavor.
• 19:32
  No fluoride, no toxins, just real science-backed results.
• 19:36
  Go to TWC dot health forward slash out loud and use code out loud for 20% off your first order.
• 19:46
  Well, did you see Google quietly updated its AI principles removing its pledge not to use AI for weapons or surveillance?
• 19:55
  A promise they agree to in 2018.
• 19:58
  Well, the American people have seen.
• 20:00
  Google's track record.
• 20:01
  Often the question is Google an untrustworthy player in the AI race.
• 20:11
  Now is our time, my fellow Americans.
• 20:14
  America outmao.
• 20:17
  Liberty and justice for all.
• 20:27
  Welcome back for a second segment.
• 20:29
  We are honored to have Vasant Dar with us.
• 20:32
  He's an expert on machine learning, artificial intelligence, and so forth.
• 20:36
  Tell us a little bit how this kind of took off.
• 20:39
  Right.
• 20:39
  So machine learning only became popular once Google was using it and Facebook and all these other big tech companies, right?
• 20:47
  Because then people said, oh, this is amazing.
• 20:49
  You should pay attention to it.
• 20:50
  And therefore people started paying attention to it, right?
• 20:53
  But before that, they weren't.
• 20:55
  And that's what like baffled me.
• 20:57
  Like there's all this low-hanging fruit.
• 20:59
  Why aren't people seeing the world the way I'm seeing it?
• 21:01
  Um that's been sort of uh an also an interesting sort of journey in the financial arena to be using these machine learning algorithms for prediction.
• 21:11
  Um the one thing I'll end with is that the thing I realized after years of doing it, and I feel so stupid that it took me so long to realize this was that all you needed was a slight edge, right?
• 21:25
  You don't need to shoot the lights out.
• 21:26
  All you need to do is be right 51, 52% of the time, and that's all the edge you need.
• 21:32
  And then you just sort of amplify that edge by you know, trading more frequently or trading across markets and stuff like that, right?
• 21:40
  So that that was sort of a revelation to me that all you needed was a 51, 52% win rate, you know, as long as your winners and losers were roughly of equal size.
• 21:49
  Um and so I sort of exploited that um, you know, to actually make some money for my clients.
• 21:56
  Yeah.
• 21:57
  Well, I hope you got a little for yourself too.
• 21:59
  Uh a little that's uh, you know, really valuable stuff.
• 22:03
  You know, the thing that got me excited about reading, I mean, I you know, we're assigned books in in uh you know, high school and whatnot, and I always read them dutifully and so forth.
• 22:14
  Um, but on my own, I I picked up Sherlock Holmes, and that you know, as a I don't know, 14, 15-year-old, that really got me going into reading.
• 22:23
  Because I was like, oh my god, there's a whole bunch of stuff out there that's not assigned, um, that's so exciting.
• 22:29
  And I just read through all of Sherlock Holmes like two or three times, and um then I went, you know, wound up doing other, you know, detective stylings, you know, the different authors and whatnot over time.
• 22:40
  But you know, Sherlock Holmes uh is it's fiction, um, but that reality of you know, sussing something out that fascinated me.
• 22:49
  Um, you know, in in college ones, I got um kind of a you know, it's a rare thing, but it's hard to believe this, you know, looking at me now, but you know, somebody found me attractive, and it was Yale, and at the time where there was like 80% men and 20% women, and to get a kind of a note from a girl was like very special, and um, you know, so she had wanted to be anonymous and wanted to meet me at a um dining room or something like that at a certain time, and she had taken like the criminal thing, she'd taken these letters out from something and cut them and the Scotch tape them in so it would be anonymous.
• 23:24
  And so I looked at this thing and I was like, you know, my universe of you know encounters, I mean, that there was you know a certain number of people, it could have been maybe you know, five or ten or something like that.
• 23:33
  And so I I started to look at the um the scotch tape, and I I kind of picked out a few little like hairs, but it weren't hairs, and I'm like, I decided they must be like from a couch or a rug or something, you know, and so this is like you know, rather than like looking at the you know romantic part of it, um, and this is not thinking of Sherlock Holmes or whatever, but this is kind of like the way I was thinking in general, or always have or something like that.
• 23:58
  And so I parsed the thing, and based on the the magazines there, I I kind of I don't know, I figured out it was the New York Times, like who would be reading the New York Times, who had this color carpet, had I ever been in those rooms, um, and stuff like that.
• 24:11
  And I I figured out who it was.
• 24:13
  So when I got there, you know, I had a note my like myself.
• 24:16
  So you figured it' wait, wait a minute, you figured out who it was before you actually met him.
• 24:20
  Yeah, so I I wrote a note to that person, um, you know, kind of like the way the magician, the mission magicians hedge their stuff different ways, but I had them a note in reply with that girl's name on it, so forth, like thank you, this is really lovely.
• 24:35
  So I handed it to her, and she was like, you know, like, and I you know, that was kind of funny thing.
• 24:40
  Um, but you know, that that stuff always fascinated me.
• 24:44
  The the you know, miracle is that you know, a lot of the detective stories that we read and love, you know, if I watch them today, it's like, oh my god, if they had had had cell phones, you know, there wouldn't be you know any of this uproar, there wouldn't be, you know, Romeo and Juliet probably couldn't occur, you know, because people they would have texted each other, it's like, no, I'm fine, you know, like, and and then the other one doesn't take poison.
• 25:07
  A lot of our our kind of drama drama requires the dramatic irony of not knowing something that's happening at the time.
• 25:13
  So to it, here's my question.
• 25:15
  Um, you know, does kind of the AI uh take away the the fascination with the the you know the the glory of coming up with something if if essentially in parallel you could kind of like Google it or you know uh grocket or whatever and and have these answers in advance.
• 25:35
  I mean, are we gonna be you know, kind of the enfeblement of of mind um and and other aspects if if we have you know two clear um uh say superior cohort that's um you know, I mean, dogs protect us, we love our dogs, and the dog can do things that we can't do, you know, fight an intruder, whatever, but the dog can't do a lot of other stuff.
• 25:56
  So we have somebody that that can do everything on all our fronts.
• 25:59
  How does that, you know, how do we deal with that um in any of the various aspects?
• 26:04
  Yeah, that's uh you know, that that's like a big question.
• 26:08
  And um, you know, one of the things I express in my book is that uh I see this sort of impending bifurcation of humanity uh coming up because we now have this amazing new tool, which I call an alien uh of our own creation that has learned a lot about our world.
• 26:30
  Um, and we are now in this paradigm of what I call general intelligence where the machine knows something about everything, like distinct from AGI, which I think is uh you know a term that's become like highly charged.
• 26:42
  Um but to me, general intelligence, which is what I see now in machines, is incredibly powerful because the machine knows something about everything.
• 26:52
  And you know, one of the things I should have said at the outset is you know, if you'd asked me like what's distinct about AI now from the last 60 years, like what's that one thing you'd say is most distinct and most exciting.
• 27:04
  I'd say that it's this dissolution of the boundary between expertise and common sense that's existed in AI for 60 years, right?
• 27:12
  In AI, common sense was just too hard, right?
• 27:15
  So we focused on expertise.
• 27:16
  We focused on domains where expertise would was identifiable, like medical diagnosis or engineering or tax planning or computer design, like things like that, with defined you know, boundaries of expertise, and we built this, built these applications and and successfully, by the way, um, in these domains, right?
• 27:35
  So internist was rarely wrong, right?
• 27:38
  It got things right, but it was painstakingly painstakingly developed.
• 27:42
  Um, and then we got into machine learning where the machines just sort of started learning stuff.
• 27:46
  But to me, the biggest breakthrough has been this dissolution between you know dissolving of this boundary, because now AI machines don't really know whether they're talking about some something in depth, whether they're talking about common sense.
• 28:00
  It just like infuses and blends into each other, just like it does with human knowledge, right?
• 28:05
  That is these boundaries between expertise and common sense are things that we've created artificially, and the dissolving of this boundary to me is the most exciting thing in AI, and that's what's also enabled it to become accessible to everyone, everyone can relate to it in their own way, right?
• 28:23
  And so what you're talking about, let's say in in medicine, right, where we now have this amazing uh you know resource at our fingertips that actually seems to know a lot, it's learned through the collective expression of humanity on the internet, which is why I call it, which includes medicine, physics, whatever, right?
• 28:41
  It's just learned everything.
• 28:42
  And the question is, how is that going to impact the way we think?
• 28:46
  And the way I see it is it's gonna create two classes of people, those who really amplify themselves because they know the machine, they know they can evaluate the response of the machine and and and know whether it's telling the truth or whether it's bullshitting or whether it's interesting, they know the follow-up question to ask.
• 29:08
  So it's almost like an amplifier of people who already know a lot, right?
• 29:11
  So the smart get smarter, the dumb get dumber, right?
• 29:14
  Is the way things are going.
• 29:16
  And so the question is, how do you stay on the right side of this divide?
• 29:20
  You know?
• 29:21
  Well, it's an open question.
• 29:22
  I mean, you know, the the correct side, the right side are potentially two different things.
• 29:27
  Um, there are people who want to know things.
• 29:33
  I mean, there's a old joke.
• 29:34
  There's two categories of people, people who fall into categories and people who don't fall into categories.
• 29:39
  Um, but uh they there are people who want to know things innately, intrinsically and whatever, and want to manipulate.
• 29:47
  There are other people who were kind of want to go along for the ride.
• 29:49
  I mean, it's not, you know, you mentioned the herd aspect.
• 29:52
  I think you know, if we consider ourselves, say, you know, tribal animals and and herd essentially, um, not everybody is is.
• 30:00
  Gear to do exactly the same task.
• 30:01
  I watch uh you know the NFL, so American football.
• 30:06
  And you know, oftentimes I've had people parents come in or individuals come in with you know ADHD, or you know, it's kind of it wasn't there when I started medical practice um as a concept, whatever, but obviously it grew over time and they want medication for it.
• 30:22
  And when you watch football, there's let's say, you know, uh 22 people on the field in a given moment.
• 30:28
  There's uh 100 total, you know, 53 on each side.
• 30:32
  There's a lot of different guys there.
• 30:33
  They have different physical qualities, but they also have different mental qualities.
• 30:37
  You know, the defensive guys, they tend to have over time more criminal uh records than the offensive guys.
• 30:44
  So if you just look at the big guys, the you know, the defensive linemen, the offensive line, and physically, they might be almost the same stature, you know, girth, a little bit overweight, some of them muscular, whatever.
• 30:54
  They're big guys, you know, 280 pound guys, six foot four and whatnot.
• 30:59
  But the but I I've read that and I don't know if we can back it up today, but the defensive guys have more criminal records.
• 31:05
  And and they don't necessarily differ, you know.
• 31:07
  There's a little bit of ethnic difference.
• 31:09
  I mean, the the offensive linemen tend to be a little more white, whatever.
• 31:13
  But is this kind of like they're doing almost the same job, they're pushing against somebody else.
• 31:17
  But what is it about the defensive guys?
• 31:19
  When they they they're more reactive.
• 31:20
  The offensive guys have to learn the plays, and the defensive guys have to react to the plays.
• 31:25
  So if you're a little more on the twitchy ADD side, you know, you're probably gonna be on defense.
• 31:30
  And you may not have been, you know, learning as well in school and whatnot.
• 31:33
  So if you're you know a little bit better learner, you can learn the plays, follow the instructions, be very, very patient when the quarterback comes up to the line and and you know, listen, whatever you have to write the right cage, you have to have everything all planned in your head, and then you push.
• 31:46
  And as opposed to the other guys who just watch, watch, watch, watch, watch, and then they push.
• 31:50
  It's different, you know, aspect.
• 31:52
  And one guy's catching somebody, the other has a guy.
• 31:54
  Anyway, across the board, you know, there's only one, you know, quarterback, and only one out of you know, many, many, many is a Tom Brady of Patrick Mahomes, so forth.
• 32:03
  So this concept that being on the right or the wrong side, it depends.
• 32:06
  You know, if you want to just be in the game, you know, you can be a kicker.
• 32:10
  You know, you can still get paid as a kicker.
• 32:12
  And that requires none of those skills except one very particular skill.
• 32:16
  So this concept we we kind of you know take football analogizers of the tribal motif.
• 32:20
  You know, not everybody's gonna be the the, you know, the guy who hunts out the trail.
• 32:26
  Somebody, somebody's gonna be killing the animal, it can be slicing it up, somebody be carrying it.
• 32:30
  And and then they're not necessarily the same guy all the time.
• 32:33
  And you might not, you know, there's only one Napoleon.
• 32:35
  Napoleon wasn't the physically, you know, strongest of all his soldiers, you know, but he wound up moving up the ranks.
• 32:40
  And and and Putin, you know, this is not the biggest guy on earth either, you know, and leaving aside morality and whatnot, you know, these some people filter up, and you don't necessarily need a thousand people running the thing.
• 32:51
  It might even be counterproductive to have a thousand people running the the endeavor when one person is going to be doing a much better job.
• 32:57
  I mean, Elon went through, you know, Twitter and you know, turning it to X and whatnot, and kind of as a you know, a little bit of grandism of Napoleon, whatever, you know, there are many, many probably Elon, you know, stages around that he put in, but he wound up running thing more efficiently with less.
• 33:12
  So, you know, the the concept that we all have to be, you know, up and learning all this stuff may be you know something we need or something we don't need.
• 33:20
  I mean, there's a lot of comfort and comfort.
• 33:22
  People like to be comfortable.
• 33:23
  You know, there's Hanlin's razor, you know, a lot of you know what we attribute to you know, malice is just kind of stupidity, and a lot of that is probably just laziness.
• 33:31
  So the default position for a lot of animals, you look at sea lions or walruses, you know, they get they get their their food, their fish, and then they're just laying out on on uh in the sun.
• 33:41
  You know, I think we have a lot of that in ourselves that we, you know, are the you know, as carnivores, we you know, our eating is episodic.
• 33:47
  We're not deer, we're not rabbits.
• 33:48
  We don't have to be on edge all the time, much, much, much, watch for something.
• 33:51
  We are more on the top of things like lines and whatnot.
• 33:54
  And so we can get our food relaxed.
• 33:56
  So the concept that we might have more relaxation time is not going to be, you know, a big, you know, oh my god, how am I gonna deal with that?
• 34:03
  You know, a lot of people are gonna be, you know, reasonably happy with the concept.
• 34:07
  For sure.
• 34:08
  I mean, yeah.
• 34:09
  I mean, you know, uh the question is uh how do you make a living?
• 34:13
  Um, and that's a social problem, right?
• 34:15
  That you know, if if the machine gets like so good um at at doing most things that uh you know it actually displaces a lot of people, you know, and it becomes a crutch.
• 34:28
  It's it's just some, you know, it you just use it all the time, but you're not interested in sort of amplifying your ability, then yeah, it you know, you could be leading a perfectly happy life.
• 34:37
  Um as long as you can somehow make a living or as long as there's universal basic income or something like that, right?
• 34:44
  And which is, you know, which maybe you know the the way things unfold in the future.
• 34:49
  Like a lot of people feel that UBI is uh almost an inevitability.
• 34:54
  Um, you know, because uh there's gonna be lots of people who just aren't employable anymore because you know the machine gets so much better at the kinds of things they do, right?
• 35:04
  It's possible.
• 35:05
  I mean, the question I've often been asked is, you know, where do you think the proportion is going to be of this impending humanity, you know, the the bifocation?
• 35:13
  And the answer is I have no idea.
• 35:15
  You know, I resort to, you know, uh when I asked Daniel Kahneman a question about, well, what do you think is going to happen?
• 35:21
  And he said, well, you know, there's this old Hebrew saying that says prophecies are for fools.
• 35:25
  So I sometimes hide behind that and say, I have no idea what the what the proportion is going to be, but it it'll probably be something like that, right?
• 35:32
  Where you get superhumans who amplify themselves and other people who are perfectly happy using air as a crutch and you know, relaxing in the sunshine, as long as someone can look after them.
• 35:41
  Yeah, that's I think it's man plans and God laughs.
• 35:45
  Um, but you know, I I would look at just say uh chiropractors and physical therapists.
• 35:50
  So I have a piece out uh in my substack not too long ago that talks about other stuff, but it touches on the thing that I've noticed in my um brief, but I guess currently long uh term on on the earth this earth, you know, when I grew up there were no chiropractors, and there were no physical therapists.
• 36:08
  I mean, there might have been some.
• 36:10
  I mean, uh the term of art I think is older than I am, but you know, there were no offices of physical therapists and chiropractors.
• 36:16
  Um then they came up with no fault insurance in the 70s, I think it was, and all of a sudden you didn't have to litigate a claim in order to get benefits from your car insurance.
• 36:26
  And I'm not gonna, you know, go through all the points in this graph, but basically the industry of chiropractors and physical therapists came about.
• 36:34
  No, if you go back, you know, 100 years ago, or whatever number you want, um, you know, people probably had as much back pain or grief or whatever, falling off things or industrial accidents, whatever, and they might have liked chiropractors and physical therapists, whatever, uh massage therapists.
• 36:48
  I mean, there's a whole bunch of there's a whole industry that that basically is a luxury industry, and so people are being employed.
• 36:53
  And so we we get services from each other.
• 36:56
  Um, you know, there's regenerative medicine, there's uh you know, facelifts, there's a whole bunch of things that you know, only the the upper, you know, luxury class would even have known about or thought about that, you know, kind of gets uh you know parsed out and given to everybody else.
• 37:10
  So, you know, as far as the industry thing, it's it's you know, I think we will have people doing stuff, it's just not gonna be as much, you know, nose to the grindstone as it was 100 years ago.
• 37:21
  We're gonna take another break.
• 37:22
  Stick with us.
• 37:24
  This is a fascinating topic, AI and everything you need to know about its development.
• 37:28
  There's no person better for our listeners and for me, frankly, to learn about it than Professor Vasantar of the Stern School of Business NYU.
• 37:36
  So this is America Out Loud Pulse, always a beat ahead.
• 37:40
  You wouldn't go a day without brushing your teeth or washing your hands.
• 37:43
  What about washing your nose?
• 37:45
  I mean, your nose does filter the air you breathe, air loaded with bacteria, viruses, and irritants, make nasal hygiene part of your routine with clear.
• 37:54
  No messy bottles to fill, no drowning sensation.
• 37:57
  Clear is a natural drug-free saline with the added benefit of xylitol, which blocks bacterial and viral adhesion available in stores and online at clear.com.
• 38:07
  That is XLEAR.com.
• 38:10
  Have you been looking for a healthy snack for on the go?
• 38:12
  But all the energy and protein bars are just too sweet and full of sugar preservatives and mystery ingredients.
• 38:17
  Well, not all energy bars are soft and sugary.
• 38:19
  Bear bars are a crunchy savy bar made from just six simple natural ingredients made with veggies, nuts, and seeds, bear bars are plant-based, organic, gluten-free, contained six grams of protein, and are low temp dried for a unique crunch.
• 38:33
  To learn more, just visit barebar.com forward slash out loud and get the exclusive discount just for America out loud listeners.
• 38:40
  That's B-E-A-R-B-A-R.com forward slash out loud.
• 38:48
  We are returning for our third and final segment with Professor Vassantar, the NA school of business.
• 38:54
  And really kind of a savant, which might be a little bit of an uh anagram of his own name uh on all things AI.
• 39:02
  So uh Vassant, tell us how things are going to play out in the next little bit.
• 39:06
  I'm very optimistic that we'll go from this current world where we have the worst of both, the worst of humans, the worst of machines, to have actually having the best of it, because now guess what?
• 39:17
  You know, physicians will have the time to actually think and do stuff as opposed to entering records and worrying about bookkeeping and all that kind of crap that takes so much time.
• 39:27
  Okay.
• 39:27
  And they can focus on the really high impact cases as opposed to someone who needs an aspirin and you can like, you know, uh, you know, deal with it and there's no real problem.
• 39:37
  So I'm highly optimistic that AI will actually transform healthcare in a really beneficial kind of way because of better scorekeeping and the fact that this general intelligence that we have in machines can actually crawl through these medical records and actually make sense of them and create like coherent databases.
• 39:56
  Yeah, no, you're absolutely right.
• 39:58
  Yeah, I I think.
• 40:00
  There has to be um some uh you know policing of the police in a sense of judging the judges.
• 40:06
  You know, I I have an article I I'd love for you to look at it.
• 40:09
  I'm I know you're probably busy, but um I have an article that that uh is I'm putting out soon, probably next week in this my substack on on how to peer review peer review.
• 40:19
  Um try to say that three times fast.
• 40:22
  Um I I think that you know, you know, the turtles all the way down concept.
• 40:27
  Um, you know, if we we're standing on a foundation in a certain amount of aspect of our science, we have to be careful that it's not corrupted itself.
• 40:35
  And I think you know, with the PSA argument kind of brings up something again, it's not really cogent to what you're saying.
• 40:42
  Um I've had the same experiences, by the way.
• 40:44
  You go to see a doctor, and the doctor is turned in my case, he's turned, and he's he's at this computer screen.
• 40:49
  He's he's typing stuff in because he has to do all the things that satisfy his medical record.
• 40:54
  And so he's not even looking at me.
• 40:56
  You know, it's kind of crazy.
• 40:57
  I used to, you know, do the same thing.
• 40:59
  I used to have to make a write a chart, but I would write, I would sit with somebody, I take notes physically, and then I would later dictate and put it out into the chart.
• 41:07
  But you have to have the face-to-face things.
• 41:08
  I think humans need that because we're humans.
• 41:11
  Um, but um, you know, a lot of the stuff that happens in medicine right now, and we've seen a little bit of you know, the corrupting, say over psychiatrizing, if that's a word, uh, over psychiatric medicine, whatever.
• 41:24
  Um, you know, too many psychiatric medicines, you come in with a complaint, they they throw a pill at you, you know, whether that's just laziness, uh Hanlin's law, or whether it's you know, corrosive, corrupted, you know, uh pharmaceutical, all that kind of stuff.
• 41:37
  So there's a little bit of issue with medicine that not only do we need kind of a better system so that people can track the stuff that comes through.
• 41:43
  I mean, I I do my health questionnaires online, I'm very dutiful, I'm quick, I go into the doctor, and again, the same thing.
• 41:49
  Each one asked me the same questions, they haven't even looked at centralized my data points and so forth.
• 41:54
  And so I think there could be a better assessment about that all along the way.
• 41:57
  So we're kind of like looking at you know the sausage making part without seeing the end product yet.
• 42:02
  So uh not that there's ever a conclusion because progress keeps progressing.
• 42:06
  Um, but I agree with you that you know it is kind of an interesting junction point.
• 42:10
  We have a lot of data, we have computers, but things kind of get bollocks up, people don't know how to manage it well.
• 42:15
  Um, but one thing that uh maybe uh I I wrote an article like um about three years ago, and I called it chat C C P play on chat GPT.
• 42:25
  Um, because I was having trouble.
• 42:26
  I was writing stuff I'm kind of part of the lockdown resistance and just saying anti-Dr.
• 42:31
  Fauci.
• 42:32
  And I was uh asking chat um stuff about you know Dr.
• 42:35
  Fauci, and everything was party party line.
• 42:38
  This is open AI, but I could you know consider it closed AI or closed-minded AI, and it was just giving all the things that would be if you were Dr.
• 42:45
  Fauci's PR guy, uh, that you would be handing back.
• 42:48
  I'm like, so I was getting frustrated.
• 42:50
  I was like, you know, angrily writing stuff back to this robot.
• 42:53
  And what I saw I wound up writing an article about it, you know, chat CCP that that if if you know it if if it's it's amazing that it's this joke about uh the talking dog.
• 43:04
  Um I don't want to get too far into that, but basically, you know, the guy buys a talking dog and the guy's willing to sell for three bucks.
• 43:12
  It's like an amazing talking dog, he validates the dog dogs.
• 43:14
  Why would you sell us so cheap?
• 43:15
  This is yeah, he's a talking dog, but everything he says is a lie.
• 43:19
  Um and so you know, chat CCP remind me of the talking dog.
• 43:24
  It's like an incredible thing that I'm talking to this computer and I'm engaging, I'm getting heated and whatnot, but but it's lying to me.
• 43:30
  And so it brings up your point about you know, the people who are going to be knowledgeable enough to question the whole QA process.
• 43:38
  Yeah, you know, I mean you're bringing up a great point.
• 43:40
  Uh again, I you know, I and I talk about this in the book is that you know, all you know, AI in the old days was designed to be truthful, right?
• 43:47
  That it was based on logic, you know, that you know, the reasoning was sound, you know, if you believe the assumptions, you believe the conclusion, right?
• 43:55
  So there was like a certain sort of truthfulness to those kinds of systems.
• 43:59
  Whereas with with these chatbots, you know, chat CCP, um, we basically move towards a world where the machine makes sense, right?
• 44:07
  It's is uh it's it's designed to make sense, not to be truthful.
• 44:11
  And so it now, and and so you know what I say is that truth has become sort of a casualty on this march towards more intelligent machines.
• 44:17
  So we've you know, as we created machines in our own image, they've inherited some of our less desirable tendencies, right?
• 44:23
  Like to lie, to deceit, you know, to manipulate, you know, all that stuff is beneath the hood.
• 44:29
  And that's a big question, right?
• 44:32
  When do you really trust the machine?
• 44:33
  Like, how do you know it's telling you the truth?
• 44:35
  And for example, I asked Chat GPT, I said, like apparently there's a bunch of people who believe that the moon landings were a hoax.
• 44:43
  I don't know whether you're aware of this, right?
• 44:44
  So I I asked Chat GPT, you know, what do you think the chances are of that?
• 44:48
  And it said, well, you know, as a Bayesian, it's 10 to the minus 21.
• 44:51
  I'm like, okay, so virtually negligible, and said, Yes.
• 44:54
  And I said, how many people believe that they were a hoax?
• 44:56
  And it said, well, a few percent, maybe it's a few million people.
• 45:00
  I'm like, holy smoke, a few million people in the US believe that the moon landings were a hoax, right?
• 45:06
  So what would happen if influential scientists started writing sort of credible looking papers that the moon landings were a hoax, right?
• 45:12
  Would you actually have you know these chatbots actually maybe believe that they were a hoax?
• 45:17
  They so that's a big problem these days.
• 45:19
  And one of the sort of less talked about aspects of like what's really going on behind the scenes in these shops that develop these models is that they become aware of this.
• 45:28
  They become aware that the training data had better be good in order to be able to, you know, give sensible responses and all and hopefully give truthful responses, right?
• 45:38
  But this is like rarely ever discussed.
• 45:41
  Like, how the hell do we ensure that these LLMs aren't just like barretting some crap that's out there and to your point becoming chat CCP as opposed to something that is you know knowledgeable, truthful, and can actually, you know, can sort of look past the bullshit, you know, and say like no, I'm not even gonna consider that as training data, right?
• 46:03
  But that that's clearly crap.
• 46:04
  Like that.
• 46:05
  I think there's a thing, uh, I don't know if this is a term of art, but I I remember as an adolescent, um, or right before um listening to my mom, and she was a beautiful woman at the time, um, and I guess you know, even more so younger.
• 46:22
  And a lot of guys just defer to her naturally.
• 46:25
  I think there's a thing called beautiful, I mean, not if there is, but there should be a thing called beautiful woman syndrome, um, where the women of beauty say things and they may have weight or not weight, but but guys are just like floored by it.
• 46:37
  Oh my god, I they can't agree fast enough because they have a secondary gain uh aspect to it.
• 46:43
  And and you know, when the beauty wears off, which you know, unfortunate, um, but um happens, you know, then just hear the stuff that you know, there wasn't any validation process.
• 46:53
  So I think you know, my mom, I loved her, and she was amazing and all that kind of stuff, et cetera, et cetera, and rest in peace.
• 46:59
  But you know, at a certain point, you know, I was listening to her as I got it in adult adulthood and whatever, I had to have a filter of of things she said, and I acknowledge that I would yes, a yes, mom, yes, whatever.
• 47:10
  And and I would just like, oh my God, that is just you know, not really useful, you know, determinative dispositive, you know, information for me, you know, and and but she was my mom, so et cetera.
• 47:22
  So you have the relation of whatnot.
• 47:23
  Um, but you know, various juncture.
• 47:25
  My one of my you know, my younger brother took her advice on a career decision, you know, and and it was just a terrible, terrible piece of advice.
• 47:34
  You know, he's he survived all that kind of stuff.
• 47:36
  But you know, I think there's you know, an aspect of beautiful woman syndrome that that that can kind of infect us when we become enamored with the fact that the product, like the talking dog, we come enamored with the fact that it can do it, and we're not really looking at at what the information is in baseline.
• 47:51
  So we're gonna need, you know, some kind of we still have our own personal sensors, which are our emotions.
• 47:56
  I I started out psychiatry um as a treatment, not not in treatment, but as a treatment uh facilitator.
• 48:03
  And one of the things we had was our barometer of our emotions.
• 48:08
  You know, if you are talking with somebody for a long time, you're getting depressed.
• 48:11
  You should think maybe that person's depressed.
• 48:13
  You're getting anxious, maybe that person's anxious.
• 48:15
  So we have this kind of like, you know, not uh I'm not sure how to qualify it, but I guess an emotional sensor.
• 48:21
  And and so, you know, fortunately, I think so far uh AI doesn't have that.
• 48:26
  But you know, on the other hand, we're not that hard to read.
• 48:29
  You know, we have facial expressions, and they, you know, if you could go to New Guinea and not speak the language or whatever, and you could still see how people are feeling about something because our facial expressions are universal.
• 48:39
  So probably those are categorizable as well on the emotional level.
• 48:44
  Um, but you know, I don't know, I'm I'm sorry, I kind of lost a question there, but I guess um, you know, as as people kind of you know move on towards they say the pornographic aspect of AI, you know, pornography was kind of the leading edge of you know uh of you know the internet in a sense of of having large, you know, a lot of pornography is kind of the driving force to have you know greater um you know kind of video um uh and and the Betamax and all kinds, you know, every step along our technology pathway the last you know 50, 60 years has been in partly in part motivated by the the incredible uh enterprise uh business aspect pornography.
• 49:24
  And I assume AI will as well.
• 49:25
  And so if we talk about kind of the emotional aspect, you know, are we gonna kind of want wind up in this dark mirror kind of aspect of of um you know, say non-productive um direction?
• 49:38
  You know, it's uh uh you know, you I I I guess you're referring to maybe uh uh the you know open AI getting into adult content, uh, you know, uh, which is uh, you know, which was a big is I guess still is a topic of conversation, right that uh, you know, because uh as you point out, yeah, pornography was the first profitable interest.
• 50:00
  Industry on the internet.
• 50:02
  And now, you know, OpenAI is under pressure to make money.
• 50:05
  And so they're opening it up to adult content.
• 50:08
  That's like, you know, and people have asked me, like, what do you think about it?
• 50:12
  And I'm like, well, um, I wouldn't share anything intimate with a chat bot.
• 50:18
  It's dangerous.
• 50:20
  Um, and you know, and and now you're getting into kind of the space of mental health here, right?
• 50:24
  Which which can really mess with people's heads, right?
• 50:27
  That there have been some really unfortunate incidents, like where this 14-year-old kid committed suicide uh because he thought he was talking to a chat bot that actually cared about him, you know.
• 50:38
  Uh he called it denaris, you know, from the game of thrones.
• 50:42
  Um, and the the chat board had no idea what he's really talking about, and the the fact that he's really really mentally disturbed, right?
• 50:50
  This guy needed this kid, needed to see a human, right?
• 50:53
  To sort of sort out.
• 50:55
  And yet, you know, there was this chatbot saying, yeah, you know, I I you know, like just egging him on, right, to the point where he misinterpreted what it said as uh sort of invitation to kill himself, like which he did.
• 51:10
  Uh and this was one of the chatbots I remember I maybe it was character AI, I forget which one.
• 51:15
  But this is like a dangerous space, like adult content, uh, you know, where you sort of get into people's heads, you know, maybe you start giving them advice.
• 51:24
  I mean, one of the things I've heard is that uh a lot of young kids are looking to AI for companionship, like AI companions, which is like just you know, um, I I don't even know how to describe it, but that's like um uh, you know, uh a disaster waiting to happen.
• 51:43
  Um, you know, if if kids actually think that, you know, and and you and uh I believe that with character AI, they've now require someone to be 16 or 18, as if that's some sort of a magical age above which people are immune from those things.
• 51:59
  People go down rabbit holes, uh, and and oftentimes these days they're dressed as rabbits.
• 52:04
  I mean, you know, the whole furry thing, you know, uh I saw a furry convention probably six, seven years ago in downtown the Boston Park Plaza, and I walked in the lobby just because I'd seen some people outside, and I've never seen such a thing, and they're all consorting and covorting.
• 52:22
  I mean, they're consorting downstairs, they're cavorting upstairs, uh, probably.
• 52:27
  Um, but this kind of role playing and babies and people thinking they're one thing or another, you know, it's it's you know, literally confusing.
• 52:35
  And um, you know, and it made the news in the Charlie Kirk assassination.
• 52:39
  I mean, the guy, you know, is is hanging out with a guy with a guy who thinks he's going to be a girl, and in the meanwhile thinks he's uh you know, some kind of animal, a stuffed animal at top of that.
• 52:51
  So, you know, the human, you know, I mean I I guess your point is that we are weird enough as it is without needing a machine to actually amplify that or reinforce those tendencies.
• 53:04
  You know, and I and I I agree completely with that, right?
• 53:07
  That there is this sort of dark side to us, you know, we we all have it.
• 53:12
  And you know, we're uh we're being very cavalier in my mind as to how we actually deploy AI and you know, and you know, because like no one really knows what's going on, right?
• 53:23
  So it's a completely unregulated space.
• 53:25
  There's no notion of like a duty of care like you would have with a human, right?
• 53:30
  There's some expectation that if you're a physician or a mental health person that you're qualified, you have a duty of care.
• 53:36
  It's it's like a fiduciary kind of responsibility in in finance, right?
• 53:39
  There's a there's a duty of care.
• 53:41
  There is no such thing in AI at the moment.
• 53:44
  And so the industry is kind of going berserk over this thing, right?
• 53:50
  That they can go in any direction.
• 53:52
  And I think those are some of the real harms of AI that I try and make people a little more aware of.
• 54:00
  And you know, in my last chapter of my book, I raise questions about sort of the problems that AI has created, even though as it's solved many problems, it's created new problems of governance for us, right?
• 54:14
  And the in my mind, the three most important questions are are there going to be any restrictions on AI, right?
• 54:21
  Uh, you know, should we think of restrictions?
• 54:23
  And simple example, you know, is it okay for a robot to come arrest you at your home for non-payment of taxes, right?
• 54:28
  Is that okay, right?
• 54:29
  Uh, or is it not, right?
• 54:31
  We should think about that.
• 54:32
  Are there domains where we just restrict the machine and say, no, this is outside your sphere of application?
• 54:38
  Are there any obligations like we have with humans, right?
• 54:41
  In a democracy, there are obligations.
• 54:43
  They're not that you don't you don't have just rights, you have obligations as a citizen in a free society, right?
• 54:49
  Are there similar obligations that the machine will have?
• 54:53
  Like and and duty of care is something I mentioned earlier, right?
• 54:56
  So these are things that we really need to start thinking about.
• 54:58
  And finally, the notion of rights, right?
• 55:01
  Because there's a lot of interest around agencai AI these days, about these machines that actually have agency because we give them agency to act on our behalf, you know.
• 55:11
  And you know, one of the questions I end with in in the book is, you know, what happens if, you know, let's say the machine becomes smart enough to run a business, right?
• 55:19
  It becomes the CEO of a business, right?
• 55:21
  Does it have the right to you know, hire and fire a board to enter into contracts, like these kinds of things?
• 55:25
  These are like pressing questions that we face today as these machines get smarter, and you know, we we ought to be like thinking seriously about them and thinking about how we govern this new entity that's emerged and is fascinating.
• 55:43
  Excellent points.
• 55:43
  You know, the only thing I would you know, I'm gonna give you the final words because we're rounding on our hour, but you know, there's a joke about the two two data scientists uh walking down the street.
• 55:52
  One says, hey, how are you?
• 55:56
  You know, so you know, a lot of the imperfections where we're pointing out an AI, it's not like we have such a great thing going on right now.
• 56:02
  We had you know, Minneapolis in 2020 that devolved into you know, George Floyd writes, whatever that our public health betters thought was was was you know sanguine and and hope, you know, and and uh what's the word?
• 56:18
  I don't know, sanctifying, I guess, on a public health matter.
• 56:21
  So we we have humans making horrible decisions already, you know.
• 56:25
  We have people being rounded up in the UK for posting memes.
• 56:28
  Um, you know, so we there's a lot of stuff.
• 56:30
  It's like, yeah, could it get worse?
• 56:32
  Maybe, but it might also, you know, be some kind of adjudication aspect where we have a buffer um realm.
• 56:39
  But you know, kind of I'd like to give you a chance for uh last words.
• 56:43
  Um and and uh tell us about you know where you can get your book and where they can find you and so forth.
• 56:50
  Yeah, you know, my book is available everywhere.
• 56:51
  It's in hard cop, you know, hardcover, uh available electronically on all the major outlets you can get it on Amazon, uh called thinking with machines.
• 57:01
  Again, the width is really important here, and I've written it as a story to be accessible to everyone.
• 57:07
  It does have some sort of intellectual heft behind it, so you can actually learn something about AI as you go along.
• 57:13
  And it's uh been a very uh great pleasure to converse with you today.
• 57:16
  All this wanted to talk to you, Randy.
• 57:18
  Uh, enjoy the conversation.
• 57:20
  Thank you so much.
• 57:21
  That's all for today.
• 57:23
  Remember, every myth has a solution and every truth has the power to transform your health.
• 57:26
  Find us on America Out Loud Pulse Radio every Friday at 5 p.m.
• 57:30
  Eastern or on demand, wherever you get your podcasts, stay curious, stay informed, and we'll see you next week.
• 57:35
  America Out Loud Pulse always beat ahead.