The Science of Smarter Thinking l Steven Pinker on AI and Human Intelligence

Welcome the world of Das a show for data enthusiasts I'm your host waren Hoffman CEO of safegraph and gplex capital for more conversations videos and transcripts visit safegraph docomo hello fellow data nerds my guest today is stepen Pinker stepen is a linguist a cognitive scientist and a professor at the department of psychology at Harvard he's the author of 12 books including bestsellers how the Mind works which I'm a huge fan of the blank state a blank slate another one I love Enlightenment now and most recently rationality what it is why it seems scarce and why

it matters stepen welcome to World of Das thank you very much now you're a proponent of the computational theory of mind which kind of treats the mind as an information processing system can you explain that a bit yeah it's a theory that attempts to explain the fundamental mystery about human Intelligence namely how can a hunk of matter be smart be intelligent now one uh popular hypothesis is that it's infused with an immaterial substance a ghost in the machine a soul a Spirit uh no evidence for that among other things when the physiological activity of the

brain stops as far as we can say tell the person goes out of existence and then many other reasons to think that the mind and intelligence inherently depends on on the brain another Possibility is that there is some uh special chemical some special substance that confers Intelligence on a human brain maybe some component of fatty component of a of a membrane or some hormone seems unlikely because no one has explained how an inert chemical can uh result in all the wondrous things that we call intelligence so the computational theory of Mind offers a third kind

of solution namely that intelligence comes from information Processing that knowledge perception ideas consist of patterns in uh in the brain that correlate with things in the world that is you open your eyes you see a dog there's a pattern of activity in the brain and it's a different pattern of activity than when you see a cat or a tree or a or a pencil uh that thinking consists of computation that is information transformed into other kinds of information by physical processes that just convert one pattern into Another so if you have the thought dog then

you also have the thought well it probably barks and it probably uh it might bite and it lifts it legs at fire hydrant at least the nail does all of these other things that you can then predict based on transforming one pattern into another and I think a third component this one is less emphasized by the philosophers of cognitive science who who articulated the theory is a theory of control of of feedback how uh You can get intelligent behavior and just as in a in a um uh a chess playing program a thermostat um uh

cruise control in a car if you've got feedback that is if you want to pet the dog if you sense that the dog is too far away to pet it if you know that walking will reduce the distance between you and the dog and to achieve the goal of petting the dog one of the things you got to do is walk toward the dog so you put control information computation together And at least you have a candidate for how matter can dis play this miracle called intelligence and when you look at the brain it's um

got a lot of the features that you would expect it information processing device to have namely its patterns can uh lead to patterns it's connected to the world through sense organs and muscles and the fact that we can simulate Intelligence on um other devices that aren't made out of brain Tissue but the doe process information namely our own computers suggests that the very idea that computation can result in something called intelligence is coherent and and our brain is is kind of in some ways who we are right like if if you and I somehow can

swap brains but keep our bodies um and you your brain is in my body my brain is in your body kind of like I would it wouldn't be the outside it'd be more the inside of who we are right well Yes no one's done the experiment but yes that's the expectation as it sometimes said this is the brain uh is this is the one kind of transplant where you really want to be the donor not the recipient right right right transplant right exactly right now how much like there's so much going on in AI right

now with large language models and neural Nets and stuff like that how how where's the analogy to the brain and where do you think it breaks down yes so They're called artificial neural networks based on the metaphor of neural networks in the brain that is many many many simple uh comp Computing elements that basically aggregate a bunch of incoming um noisy signals compare it against a threshold and then either send on a signal or not or send a graded signal you wire up um lots of them in the case of the human brain there 86

billion in the case of some of these um large language models it's on the order Of uh of a trillion or more connections that is um and uh they learn from experience in the sense and you know as do we in the sense that the connections the connection strengths between these simple computational elements that is the probability that the output will fire given the input can change as a result of being exposed to correct information that is either by a teacher and what's called supervised learning or just by statistical regularities in the Input in the

case of unsupervised learning so there's the analogy and clearly the brain is a massively parallel network of simple units um on the other hand there's some also some pretty striking differences between large language models artificial neural networks in general and real neural networks that is the kind that we have your one of them is that we don't need a uh a paby of data to as children with their child yeah that would take Literally hundreds of thousands of years of learning and kids are speaking pretty fluently by the age of three uh and also it's

best we can tell and this is uh I guess a controversy that I've taken aside and human cognition uh doesn't just aggregate statistical patterns but we also have symbols for uh for ideas for things for people for events for places for paths uh the fact that as impressive as the large language models are that they are prone to hallucinate To to mash things up to give a plausible sounding thing which doesn't actually exist in the world suggest that the fact that they don't have symbols for uh things like people places articles books uh events places

is a real handicap in these systems compared to what the human brain can do the human brain is also fairly low power right sorry the human brain yeah when it come compared to like a machine or energy or that's right well that yes That is true the uh the the energy consumption required to train a large language model is is a is is gargantuan compared to what we can get out of a cheeseburger yeah yeah exactly what what what has surprised you you or has these large has has this kind of AI breakthrough and language

surprise you at all because to me it's been incredibly surprising it has been surprising and I have to admit that part of it is that I'm kind of on Record in uh suggesting that this is not the way that the human brain achieves intelligence in particular what these models can do which uh surprised a number of us is Achieve real uh or at least a good Sim simulacrum of compositionality that is combining things in a way that the combination is not the same as just throwing all the parts in a bag um but it is

predictable given the configuration of the words so there's a Difference between a dog biting a man and a man biting a dog that's the difference between news and not news and um you'd think that just a bag of words without any syntactic rules like the subject of the sentence is the doer of the action would have a lot of trouble um especially when you embed sentences inside sentences inside sentences and especially when it comes to things more complicated than biting when it comes to something like you know the style of uh Winston Churchill yeah or

a um romantic comedy type plot um and arranging all of the compos components that actually satisfies that kind of description is something that you would think would require a lot of computation of symbol crunching and these models uh simulated without that now so that that is surprising um the question is well I guess the part of the maybe we shouldn't be surprised that we're so surprised given that there's no way that our Intuitions can keep up with the statistical patterns that are in a petabyte of data captured by a parameters uh there's a lot in

there that uh in the just in the massively high order statistics the patterns of patterns of patterns of patterns of patterns I guess we don't have any right to predict what they're capable of as a cognitive scientist I have the challenge of saying well how do we know that that's not just what the brain does yeah And I think it's too lead to say I doubt it simply because of the the combination of the sheer problem of how much data you got to train these things on compared to what a human child gets and the

flavor of the errors the bizarre hallucinations confabulations that humans uh don't do together with the things that they're amazingly good at doing you know write a limeric uh with a light bulb joke about um U George W bush and I couldn't do that but these systems Can do it so they're both smarter than humans in some ways dumber than humans in other ways probably different another thing that I I at least my intuition was completely off is the generative AI basically creating photos and videos and arts and you know all this other stuff that it

it is um and just the the rapid pace of that if you asked me five years ago that to me this would be the most um uh um uh uh or the least likely thing I would expect AI to to do right away but It but has done a really good job yeah indeed to make a uh draw a barer made out of avocado flesh uh for example it's not just you know a chair next to an avocado but the uh texture of the avocado is actually in the shape of the chair so again that's

compositionality it's combining ideas in a way that is not just putting them side by side and it's one of the classic Arguments for why human cognition should consist of manipulation of symbols there's a Difference between a chair in the shape of an avocado and an avocado in the shape of a chair uh you wouldn't think that just training a system on lots of captioned pictures would allow it to pick out that those relationships uh but it can in part because it's been designed with a specific architecture namely it does not map directly from words to

pictures but it has a representation that is NE it's kind of Independent of the senses it's not it's neither language nor is it pixels yeah it's abstract idea is and the fact that D similar systems were innately designed to have that middle level of representation I think is significant and it does vindicate one of the key ideas in cognitive science namely that we have abstract mental representations that are not the same as what comes in on the sense organs I I remember five years ago playing around with like this Word to V stuff like you

know King minus queen and the answer would be Prince or something um like kind of completely blew my mind and I it like and today that just seems so primitive even just a few years later like we are moving at a very interesting pace and at a pace it's very hard to predict where we're going to be a year or two from now well we are although as with the large language models the generative um uh graphic models do make some rather bizarre Mistakes like yes uh show me a chimpanzee writing a a column and

you've got a chimpanzee in front of with a laptop sitting on a Corinthian column it didn't get the idea that the column here first of all is only the metaphorical column and if something is mentioned in a sentence it has just one role you don't just throw in a bunch of column thoughts and mash them together uh that's something that neural networks without being structured into symbol Processing systems are prone to do this is something that Gary Marcus and Alan Prince and my other collaborators and I pointed out back in the 80s when we first

criticized these models as they were proposed as theories of human cognition for example the uh just an example we point out in the 80s and it's the same as the the the the chimpanzee writing a column on top of a column if you give it the sentence uh the um uh a bat broke a window now what we think of Is someone threw a baseball bat and it shattered a window but what the a neural network model comes up with as an interpretation is a little bat you know flapping around with wings has a little

baseball bat in his hands and it breaks the window you know not getting the idea that the whole point of syntax is that of language is that the concepts represented by words are given a role in the sentence they're the doer they're The done to they're the path they're the instrument yeah it's not just a pile of different things that they could be doing but it does seem like those things are things that one can correct for relatively quickly and we we should expect that you know the the bat broker window thing should be corrected

relatively soon no uh perhaps uh the the um and the question is will there still will it just expose other vulnerabilities so we Don't know you might be right and um the but as a and they really in when it comes to artificial intelligence and natural intelligence there really two kinds of questions one of them is how do we get the smartest possible artificial system yeah and the other is is this a way of understanding what the human brain is doing I'm naturally more interested in the latter because that's what I do yeah yeah now

you've argued that a lot of fears around AI safety Whether it's Lukowski or other types of fears are are a bit overblown like why do you think that well it's not AI safety it's rather the idea AI is an existential threat yeah so I think you know AI could do a lot of mischief hard to detect fakes for example but I think the uh the the uh imaginative scenario in which an artificial intelligence turns us into raw materials for some task that it's given paper clips and it uses us as a raw material That I

think is is uh nonsense and why do you think that like what's the well a number of reasons one of them is one thing that's not intelligence that's stupidity yeah so it's not we're not even talking about artificial intelligence intelligence consists of satisfying multiple goals A system that is uh satisfies one goal regardless of all the side effects is not an intelligent system it's a it's a weapon and if we want a solution to the problem Of how do we not get cannibalized by systems like that it's like you know don't build it don't be

don't build a system so stupid that it would pursue one goal at the expense of everything else that's number one number two don't Empower it don't plug it into the you the grid don't give it an army of robots and number three don't if if you yourself are an engineer uh you should be not so stupid as to think that intelligence consists of pursuing one Goal at all costs and empowering it to do so so that's part of it but you don't think like there's you know the the unintended consequence could be just like it

you know the system doesn't want to be shut off and so it figures out some ways not to be shut off and that could have some you know bad effects for humankind or something don't build a system that does that that is the that would be the stupidest possible system to build but I mean we have we have um You know hundreds of thousands of humans Building Systems right now um you could imagine that not all of them are going to think about all these safety things and some of them will have different goals and

some of them will have uh different time Horizons some of them just care about the next you know few months rather than the next millennium so yes I I I I you like we can't necessar rely on every person being um uh a Forward Thinking person right no But if someone built a uh started to build a killer robot uh just as if someone built a uh you know a car with a machine gun turret that sprayed uh bullets uh 360 degrees Yeah we'd arrest them for the damage they did Fast yes of course so

we're not talking about one Rogue if we're talking about one Rogue then yeah there could be a danger but there's an awful lot of people who are want to prevent that from happening yeah I think there's a mistake in assuming in In anthropomorphizing uh different kinds of intelligence and assume that if something is welld designed to solve problems it will also be interested in self-preservation now those things come bundled together in us because we're products of natural selection that's that's how we were built yeah but anything that we design will will will carry out whatever

goals it's program to carry out uh and there's no inherent Reason why uh being utterly Unstoppable preventing any external control should be built into a system that's an idiotic thing to build into a system uh it's the kind of thing that natural selection might build into US yeah but uh just because a system is really really smart that is I don't I don't see chat GPT trying to take over the electrical grid and nuclear um missile systems and uh and even if it tried there's an awful lot of Infrastructure to prevent it from doing so

now how do you think of um you know if if we just think of our own cognitive abilities when we think of all this new technology whether it be social media the internet Etc do you think it's having a very positive effect on our cogni abilities a negative effect no effect well it's having you know some of each certainly um all forms of computation including uh all the kinds of artificial intelligence that we've Been taking for granted for decades uh we tend not to call it artificial intelligence once it works very well and it's embedded

in our systems but um so just to give you an example um when I started teaching intros pych and I wanted to impress students and what an amazing thing the human mind is I said uh no AI system can recognize printed text as well as a human uh and I showed them why they're different type fonts and sometimes they're blurry and so on Uh well I had to get rid of that part of the lecture because you could buy a you know $100 printer scanner at uh from from Amazon and it can convert printed text

into asking characters no problem uh and you know likewise a lot of things that uh we were going to take are going to take for granted are already AI in our lives now like Siri and so on uh and there's no question that just as calculators made people better at U uh overall at and computers at at using Math and retrieving information from data sets and navigating our way across City um there's lots of scope for the human brain to be much much better if it's augmented by these assistants that can do things that are

smarter than we are in certain respects no but the internet you know it can you know there's always a danger that if you make something too easy then people will not do the due diligence of Discriminating bad advice from Good Advice uh that uh there'll be ways in which it can be put to use to empower um malevolent actors such as in generating uh fake news or dis and disseminating fake news that there's such certain features of the incentive structure of social media that bring out the um the the worst in human discourse instead of

the best you know such as the easy ability to um leave a snarky comment or to yeah someone as a bigot uh With no consequences for yourself so you know any system is going to have a lot of causal effects and it would be a miracle if all of them were good that just never happens and it's just a question of monitoring the good effects and bad effects and then instituting workarounds to minimize the bad effects now this is a data podcast um are there certain like data sets that you wish you would you could

have access to that would really you think help move the Science help move Humanity yeah I think U the more the more data the better and higher quality data the better that is I I've often said that people should um spend less time reading the New York Times and more time reading OurWorld and dat.com uh because journalism is a systematically misleading picture of the world because it's a non-random sample yeah of the goriest flashiest loudest it's the man bites dog it's it's man yeah it's man Bites dog and it's uh um you know even dog

bites man as opposed to the huge number of dogs that never bite anyone right in fact even if we had nothing but stories of dogs biting men that would be bad because then you think that every dog is dog is bad right yeah exactly and that is true of uh violence where um uh certain categories of violence like school shootings or terrorism which have a pretty small Death toll compared to um many other ways in which you can get killed including violent ways you can get killed like an argument over a parking spot um and

uh uh police shootings um uh hate crimes compared to just police blad homicides uh domestic violence which kind of trickle in constantly but never make the news we misled about what is safe and what is dangerous likewise when it comes to um uh say climate change we don't have a running meter in the Newspapers of CO2 emissions and of which countries do better do worse how does it change when a c country introduces or changes a policy uh and so the trends that actually shape the world are missing from most people's largest source of information

namely the news there there is something in human nature to catastrophy I to kind of think the world is ending like this has been like throughout history and throughout time humans have gravitated To many of these stories what do you think is so innate in us that we really gravitate to those stories so much one is that um there there's an asymmetry in uh the how things can affect you uh owing to the fact that we're extraordinarily complex improbable systems that it doesn't take much to uh make our whole system shut down y uh a

bit of poison a a rock to the Head um uh where so there are lots and lots and lots of things that can go wrong there Are very very few things that could go right in comparison I mean if you think about what are all the terrible things that could happen to you tonight before you go to bed it's a pretty long list yes what are all the wonderful things that could happen to you before to go you go to bed it's you know hard to think of a whole lot of them yep just the

nature of existence the tragedy of of of of human existence is one part of the answer so we're naturally more Vigilant against things that could uh harm they're more of them and they can do us more harm uh but also we um did not evolve in a world in which things progressively improved that is the the whole idea of science driven progress of Enlightenment driven progress uh that's kind of that's recent in human history and the the very idea that we can collectively put our heads together and Conquer disease conquer hunger Maybe even conquer war

is just not a uh it's Not an idea that ever had any validity now it has some uh but it's not intuitive yeah in your book the better angels of our nature you kind of put forward put forward like a theory of declining violence in the world and where we've been on a you know a general positive trajectory over the last few hundred years with with some blips along the way um but there there you know the counterargument is you know people say okay these things that are simmering and Maybe there's some analogy of being

the turkey in October not knowing it's going to be Thanksgiving in a few weeks like what do you say to those critics yeah well there is it is it is certainly true that there is um that a decline in violence doesn't mean a disappearance of violence it is also true that um violent event that there there could be nasty surprises there often have been there the none of the curves of violence or of um human Progress in general are uh monotonic no straight line right well they're not they're not they're never a straight line they're

not even a smooth curve they got they've got Jags and and spikes and no we don't know that that something even worse is going to happen but the in in the future but the the mo thing about the the turkey before Thanksgiving is is absolutely the wrong way to think about it that is it's not something that is inev that a catastrophe is not something That is inevitably going to happen yeah that you wait long enough we have agency essentially well we have agency also the world is probabilistic there is no uh executioner that is

looking at his watch waiting to blow up the world uh the way to think if you want to think about it intelligently you've got to think about the fact that catastrophes like Wars are uh random in time what are called a Pon process that is every moment of time there's a certain probability that they Occur with very little memory that is it's not the case that the longer you wait the more likely a war is to happen happen uh as best we can tell from the statistical distribution of Wars so it's not that the world

kind of builds up a lot of uh tension then has to be discharged and the longer the period of Peace the higher the probability of War doesn't work that way which which is maybe true with like recessions and stuff like that at some point you start To see these recessions in the economy right after well actually that's that that's not so clear and I I mean I'm not an economist but I I suspect economists would disagree with that that they can occur at at this so-called business cycle may not really be a cycle but it

might just be bad stuff can happen at random times and uh if you go for a long period of time without one it doesn't mean that it's like an earthquake that the tension is building up and it's Going to release my understanding is that's not a good model of the economy or of Wars uh but rather it's a game of roulette maybe it's a game of Russian Roulette yeah that is you're taking a chance at every moment of time but what it means is that it's not a question it does not mean oh this is

terrible we've had so much peace we've had so much Prosperity the big one is bound to happen that's not the way it works uh what the way it does work as best we can Tell is that it is like Russian Roulette but we can try to have many many many more cylinders in the pistol yeah and you know maybe we can have blanks instead of real bullets or uh you know we we can change even with a random process over time a pant process it can be non stationary in the sense that the probability can

in in principle change over time that's what we ought to be aiming to do namely make it less and less Likely that a big war breaks out now you've also been and and oh sorry I'll throw one more thought and if one does break out to see to it that the damage is uh as low as possible so both the timing and the magnitude yep now you you you one of things that you've been a big kind of proponent of is kind of this idea of rationality um how do you advise people to have a

more rational approach to things like Risk yeah um partly it's to be aware that our um system for assessing risk that is wired into US is not the best that we can do uh we evolved in a world that did not have data or statistics or recordkeeping agencies or a causal understanding of the world the best that we could do is um hear say stories from other people of what can go wrong and our own experience seeing things that can go wrong and getting a sense of how often they occur as a proportion of the

Amount of time that's passed that's better than nothing but it's worse at least for rare um disastrous events like car crashes and plane crashes and um or or for that matter things that um uh falling off ladders drowning we're they're rare enough that our own experience is going to be a poor guide yeah and we're fortunate enough to live in a world that has gather data and we're better off knowing what the the uh how the data work um so be aware of the Availability bias that is something that pops into mind is going to

be perceived as more likely you read about a shark attack um and so you don't go into the water but you don't pay attention to the risk of driving down the highway to get to the beach where you're much more likely to be killed but those car accidents don't attract as much attention um you also uh should be we should all be aware of biases like the mys side bias that is if some belief is A u to your tribe or G your tribe exactly yeah that is if something is a sacred value of your

political party your religion your social class you're likely to believe it with very little evidence even with contrary evidence um you're likely to look for evidence that confirms your beliefs and not that that disconfirms them so the number of biases that one should be aware of one I think it would be good if in our general Discourse in journalism in in um education we had a scale to calibrate relative danger someone suggested micromorts that is yes one in a million probability of dying in a year just so we know how much more dangerous or safer

something like nuclear power is compared to Coal Fired plants or driving compared to Flying comparing compared to biking uh anyway that those are couple of ways of being more risk Savvy and how do you But even like I I consider myself a fairly you know data oriented person but you know even when if I'm in a airplane and there's some turbulence or something I get a little restless and even though I know intellectually that the plane is unlikely to to go but you you you start to color it with your own fears your personal experiences

I do know one person who was in a p you know a well you know passenger airplane that crashed um Luckily he survived but like you start to you start to color it with all these other kind of experiences like is there any like kind of simple way of getting out of that or is that just what it is to be human uh it is what it is to be human I think well there but that having been said one can we do have these big frontal loes that can um can't turn off our emotions

yeah that is if I was you know dangling from a safety harness you know 50 stories above the ground yeah be Pretty scared was perfectly sound and the harness was perfectly you know I'd still be terrified right you got that you got to accept uh you know on the other hand you do in making decisions especially proactively uh Should I fly or should I drive where you're not in in the uh the throws of the momentary passion I got it that's a little bit easier right that makes it easier and even when the time comes

there is a lot of wisdom to be had in you know Considering the reality and not surrendering to the emotion yeah it's kind of one of those things where if there was someone next to me freaking out I could probably calmly tell them that there's nothing happening but inside I would be freaking out too exactly yes uh now you've also been kind of a a vocal critic of the decline of intellectual diversity in universities and academic freedom you recently created this Council on academic freedom At Harvard to address this can you explain like why this

movement is coming out of Academia oh because well two reasons one of them is that Academia ought to be the institution in which ideas can be freely exchanged um that's what Academia is you have this Veritas which is almost in every University right well especially my employer har that is emblazoned on on our are uh are are branding everywhere very to us um so it's and because we are Subject to cognitive biases we think we're infallible but we're not uh we think we're omniscient but we're not that's why we set up institutions like scientific societies

and a free press and Academia where ideas can be expressed and then evaluated because um no one a prior knows what the truth is so if we want Fair tasks we've got to be able to get the ideas out there and and uh go after them if people in power get to criminalize opinion get to say well If you express that opinion uh we're not going to let anyone else hear it or we're going to punish you you're disabling the only mechanism our species has for approaching the truth namely conjecture and reputation uh putting an

idea out there seeing if it survives U so Academia ironically should be the place that that is uh just built on on intellectual Freedom it seems that Academia is the place that's like most under threat uh like why why is that Like why like I I would have if you had asked me at some point in the past I would have said yeah maybe Academia is one of the ones that's suceptible to this but there'll be other institutions that are more likely to be under threat before Academia yeah um it is a a pathology that

somehow got built into the modern University and it's not exactly um clear why probably several things but but many I have found that many academics seem to Think that their mission is not to be part of a collective effort to approach the truth but rather to advance the right moral values the right uh to to to to to uh um achieve social justice now I think we just because we have a PhD that doesn't really equip us to be moral to the world that's not what I was trained to do it's not what our institutions

are set up to do but I think many academics have that conceit and from there it's easy to um uh use differences of opinion Given human nature that we always think that we're virtuous as a kind of bludgeon to um uh to to denigrate uh uh social and class and cultural Rivals and we haven't probably there are features of human nature that just make it natur natural to assume that we're right and that our enemies are wrong and that unless you explicitly build safeguards in to prevent that from happening people in power will exercise it

just as religious authorities and political Authorities have since time immemorial you give people some power you don't curb that power with open debate open criticism protection of uh of critics and it will I think naturally regress to a system of intellectual repression well it do there are some like facts that one you know somehow can get imparted through a lecture um especially in science these are like real facts this is how certain you know elements bond with one another This is certain mathematical formula there are certain facts that that happen and in in some ways

those are easier to teach because um you know you can impart them and someone can learn them and then they're there are facts that are you know that are very hard to disprove um and then there are other things that are more ideas um and you know where you can have lots of very very smart people who disagree and in those cases you really want to give People samplings of the different things so they can you know come up with their own internal truth of it but it seems to me like the people who teach

the ideas kind of also they they kind of think of themselves sometimes as the people who also are the ones who teach the the the real facts yes now there's definitely there's definitely a tendency toward overconfidence in one's beliefs and and a kind of Moralistic um bias to think that one is oneself is uh Noble and honest and uh ones people who disagree are are are are fools and are uh deceptive and dishonest and malevolent uh and we got to remember though that even the what we now recognize as pretty wellestablished facts you know in

their day they were pretty controversial correct and and lysenko and Scopes would would attest yeah and that's why it's Essential to um you know not to uh kind of give a pass to less established ideas and say well there it's really dangerous if you believe the wrong one and uh it's any idea on way to becoming established had to be debated in its own time and and uh squashing the debate had bad consequences as in the failure of Soviet agriculture under the influence of lenko who who denied darwinian Evolution now there are some um uh

uh tropes uh in universities or where you know sometimes When people say when people use the word science and quotes that it often um it often means opinion uh and you know so if you have political science or if you have um social science or something like that and if it it actually is science it's just called math or chemistry and there usually doesn't have the word science um next to it and people often use the word science to um uh put some sort of imprint on on something that they want to do like how

Do do you think the word science is being disused well um you know and I have to that has sometimes been said about my own field of cognitive science cognitive science so that's good point yeah and of course we've got Christian Science was kind of maybe a warning to us all that just calling something science doesn't make it science uh yeah but you know the term social science has been with us for I don't know 150 years or how long has It been around was it August K I forget who I have no idea that's

very interesting but it's and political science and uh it is at least an aspiration uh and I don't think anyone is in real danger of thinking that political science is a science in the way that that physics is uh and and but we also have to remember that you know physics when it comes to certain parts of physics uh the the the debates the controversies the intrusion of personal Conviction and personal opinion is as bad as in you know Freudian psychoanalysis when you ask physic about String Theory uh tempers can run high and how do

you think about about like writing for General consumption and you know I remember 20 years ago or so reading how the mind works and I I just like I love that book it really just like took me to a new place and but it was it was very accessible even for somebody like me that doesn't have the You know the Deep Research In This Place it was it was uh it was it was a relatively easy book for somebody like me to comprehend it wasn't like a total slog and and that's part of the reason

I like it so much and the part of the reason I appreciated it I I assume there's this very deep literature that uh you know somebody with a PhD might appreciate more or something um how do you think about like how should these like writers write more for the general Audience yeah so I think we we do have an obligation to uh as they as they say give our science away that is taxpayers pay for it yeah they should get it get it back in the form of understanding of discoveries and and new ideas uh

not everyone U we all have we all specialize and allocate our time to different ways I've devoted a big chunk of my time to making ideas accessible for me I find that trying to make my ideas access or ideas from my field accessible to um Smart people in other walks of life and that by the way is is what I consider my audience yeah and there was advice I got from an editor when I started out and she said academics try to write for a wide audience they often have it totally wrong they think they're

writing for for for children for for truck drivers for chicken Pluckers and they talk down so that's the wrong way to think about it the way you think about it is you're Writing for someone who's as smart as you are but just happen to go into a different line of work your college roommate who went to med school or became a lawyer or or you know an engineer uh don't condescend but don't assume that they have lived the life that you've life yes and I think the main I I've written about this I wrote a

a writing style manual based on cognitive science called the Sens of style which I identified what I thought Was the main impediment to clear writing and it isn't as most people think a desire to um use high futin jargon to to try to impress or bamboozle to show off how smart you are I don't think the motives are as nefarious as that I think it's just it's what economists called the curse of knowledge namely when you know something it's very hard to imagine what it's like not to know it it becomes second nature you don't

what what is a everyday term to you is technical jargon And you forget that you had to learn it you know way back in dad school or from some technical article and that a person even a brilliant person just has no way of knowing what that jargon means or you describe an experiment a concept without spelling out concretely what actually happened what do you what do the person in the experiment actually see uh physically and in describing things in abstractions and in technical terms you can leave out readers who are perfectly Capable of understanding it

if you just did the work of remembering you had to learn it you know when you started out and so do other people who didn't follow in your footsteps yeah in some ways the jargon is a um um it's it's it's almost a way of excluding folks from from not being in the club because there's certain acronyms or certain other types of things that uh that that are nice because they're shortcuts to get out the idea um if you say a sharp ratio to an Investor they may um have a very good understanding of what

that is but there might be a very very smart person who could understand it very very very quickly it's not like that complicated of a thing but they but they but they use the shortcut instead of like the two two more sentences to help that person along it's and it's a bad trade-off that is you the writer are saving yourself you know 10 keystrokes yeah and you are baffling hundreds thousands tens of Thousands hundreds of thousands of readers in order that you save you know four seconds in in in typing out an abbreviation you as

uh I have some you know I have some criticisms of the Strunk and whites classic handbook the elements of style but one thing they they point out that I I quoted is uh new babies are being born all the time and they don't come into the world knowing what uh uh you know priced income ratio means yeah they everyone has to learn at Some point uh now a couple personal questions you grew up in Montreal which is uh certainly one of the more bilingual places uh did that influence your trajectory into into Linguistics you know

I have to confess I wish I could say that because it's a really good story yeah the problem is it's not true because when I grew up Montreal was not particularly bilingual it was all friends yeah well it was it was um uh segregated there was an English minority there was a French majority and uh a famous book about the situation in in Quebec at the time was called two solitudes um one of the Innovations least symbolic of Pierre Elliot Trudeau father of the current prime minister Justin Trudeau was he wanted to make bilingualism a

reality at least in in the in the government you wouldn't think that would be a radical idea but they there really Was an awful lot of segregation and I regret the fact that I'm uh rather clumsy at French despite growing up in Quebec it's rather an embarrassment interesting and and you know with with these new langu large language models like in some ways you could see a scenario where translation is easier in the future and learn you know the the practicality the need for learning other languages might go down though there still might be some

beauty In learning it and you know it's just like you know learning how to play the guitar learning how to ride a horse or something like that it might be there's a lot of Beauty in it but it might not be as necessary um as it was in the past like how do you think like the just the acquisition of new languages will change in the future no it's a really interesting question and and we're already partly there we all use Google Translate we all have a a little button On the top of the the

web browser uh and it's been a huge Boon um it's you know the translations are not always accurate enough for especially if you have some sort of technical um but I Surly with a menu you know like just the ability for me to order in Spanish today is is much easier than it was fantastic yes uh so yeah I think it'll be a good thing what what it will do to the you know the need to study a foreign language in um in in in a University is is an interesting question and on the one

hand it seems rather Philistine to say oh you don't have to learn you know French or Spanish anymore and we may not get to that point especially when it comes to just ordinary conversation even there there are apps that are getting better at translating speech in real time uh but you as with all technological improvements if you can identify some good that it delivers uh And the good outweighs the harm then it feels uncomfortable but we'll probably all get used to it now do you consider yourself more of an Insider or an outsider you know

in some ways like you're a professor at Harvard that's like stereotypical Insider but you're kind of an Ayo clust as well like how do you think of yourself I I try to use my Insider uh privileges you know having tenure at a fancy schmancy name brand University to Uh promote the kind of Outsider view that I can get away with yes now I'm not invulnerable tenur professors get fired yes and that's that's part of the problem uh and I don't just say everything spout out everything that comes crosses my mind right that's crazy nobody does

that that's no no one actually says the truth all the time yeah they don't in fact my my next book is going to be about that why we deplore Uh hypocrisy and gential uh euphemism and inuendo but can't live without it oh interesting tell us more because that that I find that very fascinating so like like a lot of my books it grows out of a few observations that that I uh kind of were buried in an earlier book and I had a book called called the stuff of thought language is a window into human

nature and one of the chapters was on a phenomenon that linguists have known for decades but never really Explained which is a lot of the time maybe most of the time we don't actually say what we mean we don't just BL out the truth uh we uh we use politeness and euphemism uh if you could pass the salt that would be awesome if you could pass the salt that would be awesome what said why do we say something like that but everyone knows what it means yeah yeah just say you know give me the salt

see rud imperious there's a lot of that in in bribes in Sexual come on in compliments in criticism uh why why don't we just say what we mean uh we all think that we want people to say what we mean but if they did we would be horribly offended and uh I think the difference is and I'll here's I'll I'll I'll be brief what the book is really about is the concept of common knowledge in the logicians sense of everyone knowing that everyone knows something yeah which is different from merely everyone knowing something So if

I if you and I know something that's not the same as if I know that you know it and I know that you know that I know it and you know that I know that you know that I know you know those are very different turns out uh in particular that common knowledge allows for coordination it allows us for to do something that benefits both of us even if it's arbitrary as long as we're on the same page you know using paper currency driving on right um being Common Customs yeah common Customs being friends being

lovers being a boss and a subordinate being business transactional partners uh all of these depend on common knowledge sometimes we don't want to change a relationship uh flip from one kind of relationship to another and so we actually try to avoid common knowledge um if there's you know a couple of people who may have a sexual interest but neither one is sure that the other One does uh if you just blurred out would you like to have sex with me well then it's out there and they can't have a relationship as platonic friends ever again

or as supervisor and supervis ever again on the other hand you want to you know go out sometime you know maybe come up see the view from my apartment there even if both sides know what the what offer has been tendered ah there's a there a there's a common deniability there and there's in Particular there's a deniability of the common knowledge that is you know even if you're not naive you really know what come up for coffee means but you don't know that the other person knows that you're not naive uh and with each kind

of iteration I think that she thinks that I think that she thinks there's more and more uncertainty so that it means that you don't have to acknowledge that you're no longer just friends that there's that sexuality is on the table You can uh go back to your relationship if the offer is is turned down uh and there's there's a lot of that in not just in our everyday emotional lives but on the world stage is Taiwan a sovereign state of course it's a sovereign state but the official policy of our government is that it isn't

and of the United Nations um there does Israel have nuclear weapons well of course they have nuclear weapons why don't they just say so well there's a difference between Having nuclear weapons and saying you have nuclear weapons and having nuclear weapons and not saying you have nuclear weapons so it plays out in all kinds of fields it plays out in in um speculative bubbles why do you buy crypto well because everyone else thinks that everyone else is buying crypto y why do you sell crypto well everyone's selling it and I don't want to be they're

rushing for the exits I don't want to be left holding the bag uh so a lot depends On your thoughts about other people's thoughts about other people's thoughts you know there's this book that Sam Harris wrote like 15 years ago a short book called onling I don't know if you've ever read it uh but it basically it basically takes the uh argument that one should never lie at all and I remember reading and being like oh this is great I'm GNA try it like and I started to realize like wow I lie a lot like

about the most random things again It would be awesome if you pass the salt like it's kind of a lie like it's really not that awesome if you pass the salt uh oh there there's there there are estimates of how often people lie I think it's at least two lies a day for typical person yeah I mean probably if you're talking to a lot of people I would imagine it's way way way higher it kind of depends on how you define what a lie is but I imagine it's much much much higher uh and and

I just like the the Problem after reading that book is I actually started noticing the lies um and and it would just be like so frequent that I would say these things like why do that doesn't make any sense that I said this thing um and then frankly at some point I just had I felt like I had to forget it because it just wasn't human uh to uh to to always kind of tell the truth no and and there are some comedies that there's a there's a film called Liar Liar yes that that Great

movie but indeed uh and I think there's a reason for it and that is that there's a when you say something you're actually doing two things you're conveying information you're also kind of renegotiating your relationship and and sometimes those work across purposes we don't want to renegotiate a relationship yeah I remember I was with one of my kids when they were about three in an elevator with a a very large overweight Man and you know my three-year-old asked him just in the elevator just like hey why why are you why are you so fat um

and it was like literally one of the most awkward where I couldn't leave the elevator and just like but he didn't know any better he was only three years old like but that's what three-year-olds do not you know 30y olds exactly that's I mean that is childhood you know innocence it makes children endearing yeah but and the but the interesting Thing is he wasn't saying anything that everyone didn't already know correct knew that he was fat too yes and and and he he probably did as well yes and and the fat man knew he was

fat as well but it does change it when you blurt it out it really does and I think it's because it it creates common knowledge that is you you can't deny that the other person knows that you know that they know that you know it interesting all right last question we ask all of our guests what Conventional wisdom or advice do you think is generally bad advice well so I I guess trust your gut yeah I think trust your gut is bad in general I think that's bad advice yeah and why why do you think

so because we're full of self-serving biases and fallacies and our our gut delivers uh um advice that probably on the whole was uh a good way to propagate your genes in in the past at least for your ancestors to do it but that doesn't necessarily bring About the greatest happiness and well-being in the present I have found that at least when I when I've evaluated my own quote unquote cut uh that it it seems to work well when it tells me not to do something but it doesn't work well when it tells me to do

something so if it says don't hire this person I feel like I should listen to it but if it says hire this person right now you know or don't invest in this company seems like it works pretty well or if it says Invest in this company it seems Le less good to work um do you have any Theory as is if that is reasonable or not interesting I had I had not thought of that I I would not be surprised if that was true if if it is true uh the problem is we never know

that it's true because we remember I don't like write it all down yeah let's imagine if someone did let's say someone uh you know uh kept track of all of those decisions and then tallied up the ones that were uh uh Misses versus false alarms if it was true I can imagine one possibility is that there are a lot of ways that things can go wrong um and therefore it takes less to be able to pick up pattern match those exactly like the tiger hiding in the bushes or something or whatever yeah whereas there's it's

much it's much less probable that there' be the magic combination of features that make something succeed and so there it would Depend again this is all hypothetical uh it would depend much more on aggregating lots of probabilistic cues as opposed to one deal killer or or blackball or veto um that the aggregating lots of cues is something the human mind is not very good at doing I mean at the way that uh uh Paul m a famous psychologist put it is in terms of talk of of of the Li itations of human decision- making if

you're the supermarket checkout line you wouldn't throw all your groceries on the Belt and sell the cashier looks to me like it's $734 do you think that's okay yeah yeah yeah say no no you really have to count and when there's lots and lots of things that influence the decision one way or another you really have to count it does seem though that like there's certain decisions that people make with different time Horizons and if you're making a decision with like a subse in a subc time Horizon like probably gut is The overarching thing on

making that decision that could be if you're a video game player if you're playing basketball or something and you have to make these very very very quick decisions if you're very witty and you're making this like super quick joke or something there's probably a gut reaction there whereas if you're making decisions that take many weeks to make or something um if you have more time to make that decision which is probably true in your life Probably my guess is you're you're an academics you probably have an opportunity to to to have a longer time to

make decisions than that other people then maybe your gut is less valuable in those types of decisions no that's certainly true and it it enters into the um concept called bounded rationality namely your rationality has to take into account uh its own limitations in terms of how much data you need how much time it takes to make a decision and often The optimal way of coming to a decision if you had all the time and all the data in the world is not particularly rational if you have to make decisions in a fixed period of

time with a limited amount of data yeah Herbert Simon is the the the the man where do that out uh this has been great thank you Stephen Pinker for joining us on World of Das it's been great I follow you at saap Pinker on Twitter I definitely encourage our listeners to engage you there this Has been a lot of fun thanks so much aurin it's been good it has been good fun [Music]