0:00
Hello everyone and welcome back to Conversations with Tyler.
Search this video
0:02
Today I'm delighted to be chatting with Stephen Pinker.
0:05
Stephen needs no introduction per se, but I will point out he has a new and excellent book called When Everyone Knows That Everyone Knows dot dot dot.
0:15
The subtitle is Common Knowledge and the Mysteries of Money, Power, and Everyday Life. Stephen, welcome. Thank you.
0:25
How important is the reasoning behind common knowledge?
0:28
So say you get a bunch of men in the room and at the end of some thinking one of them concludes that he has spinach on his teeth using common knowledge.
0:38
Let's say that mode of reasoning was somehow not allowed to us.
0:42
How much lower would GDP be?
0:45
Uh it would uh it it would it would tank.
0:47
It would be almost non-existent just because all of our uh economy depends on coordination on people uh doing things that are in everyone's interest.
0:57
If everyone does them the same way, beginning with money itself, with currency, why do we accept a a piece of paper in exchange for something of value?
1:06
Well, we know other people will accept it and give us something of value. Why do they do that?
1:10
Well, they know that still other people will.
1:12
Everyone knows that everyone knows that money uh has value.
1:14
That's what gives it value.
1:16
I mean, as long as they do, and that can dis that can unravel when you have hyperinflation or a a bank run or a currency attack. uh political power.
1:24
The uh you no government can have, you know, snipers on every rooftop and telescreens in every room, you know, surveilling you 247 with the, you know, theassi ready to pounce on you if you break a law.
1:38
A lot of our our laws, a lot of recognition of authority, uh who's the boss, uh depend on just common knowledge, namely um he's the boss or she's the boss.
1:50
She gets her way because she knows I'll obey.
1:53
I'll obey because I know she gets her way.
1:55
Uh and for as long as that lasts, as long as that common knowledge lasts, you have structures of authority and reporting and power and responsibility there.
2:06
Again, it's uh it since it depends on nothing but common expectations, it can unravel as when you have a big public demonstration where people see each other seeing each other seeing each other opposing the regime and then the authority of the regime can evaporate.
2:24
But surely there's a difference between coordination and common knowledge.
2:26
I think of common knowledge as an extremely recursive model that typically has an infinite number of loops and most of the coordination that goes on in the real world is not like that.
2:36
So if I approach a traffic circle in Northern Virginia, I look at the other person, we trade glances, there's a slight amount of recursion, but I doubt if it's ever three loops. Maybe it's one or two.
2:47
And we also to slow down our speeds precisely because there are not an infinite number of loops and we coordinate.
2:54
So what percentage of the coordination in the real world is like the traffic circle example or other examples and what percentage of it is due to actual common knowledge? Yes.
3:02
So common knowledge uh you in the technical sense does involve this infinite number of infinite love of arbitrarily embedded beliefs about beliefs about beliefs.
3:13
So thank you for introducing the title with the three dots dot dot dot uh that because that's what signals that common knowledge is not just when everyone knows that everyone knows but when everyone knows that everyone knows that and so on.
3:28
Now the the answer to your puzzle and that's I I devote a chapter in the book uh to what common knowledge could actually consist of and and I'm a psychologist.
3:37
I'm not an economist, a mathematician, a game theorist.
3:39
mathematician, a game theorist. So foremost in my mind is what's going on in someone's head when they have common knowledge and you're right we don't uh we well we couldn't think through an infinite number of I know that he knows thoughts uh and our mind starts to spin
3:54
when we do you know three or four but instead common knowledge can be generated by something that is self-evident that is conspicuous that's salient that you can witness at the same time that you witness other people witnessing it and witnessing you witness it witnessing it and that can grant common knowledge in a stroke. Now it's
4:12
common knowledge in a stroke. Now it's implicit common knowledge that is uh one way of putting it is you have reason to believe that he knows that I know that he knows that I know that he knows etc even if you don't literally believe it in the sense that that thought is consciously running through your your
4:29
mind and I think there's a lot of interplay in human life between the this recursive mentalizing that is thinking about other people thinking about other people and the intuitive sense that something is out there and therefore people do know that other people know it even if you don't have to consciously work that through. And a kind of proof
4:49
And a kind of proof you gave the example of u recognition of uh well of of uh norms and laws like who yields at an intersection.
4:58
Um the uh the eye contact though is crucial because I suggest that eye contact is an instant common knowledge generator.
5:07
You're looking at the part of the person looking at the part of you looking at the part of them.
5:11
And so you've got kind of instant granting of common knowledge by the mere fact of making eye contact, which is why it's so potent uh in in human interaction and in often in in other species as well where eye contact can be a a potent signal.
5:25
But there are even species that can coordinate without literally having common knowledge.
5:31
And I give the example of u the lowly coral which presumably not only has beliefs but doesn't even have a brain with which to have beliefs.
5:40
But coral have a coordination problem.
5:43
They're stuck to the ocean floor.
5:45
They've got to uh their sperm have to meet another coral's eggs and and vice versa.
5:51
Uh they can't spew eggs and sperm into the water 24/7.
5:54
It would just be too metabolically expensive.
5:58
So what they do is they coordinate on the full moon on a particular on the full moon or depending on the species a fixed number of days after the full moon that's the day where they all release their gametes into the water and which can then find each other.
6:11
Now they of course they don't have common knowledge in knowing that the other will know.
6:15
It's kind of implicit in the logic of the their solution to a coordination problem namely the public signal of the full moon which they over evolutionary time it's guaranteed that each of them can can can sense it at the same time.
6:35
And and and indeed in the case of humans we might do things that are kind of like coral.
6:38
But is there some signal that just leads us to coordinate without thinking it through?
6:42
The thing about humans is that because we do have or can have recursive mentalizing, it's not just one signal, one response, you know, full moon uh you know, shoot your wad.
6:56
It's um there's no limit to the number of things that we can coordinate creatively in evolutionarily novel ways by setting up new conventions that allow us to coordinate.
7:07
But I'm not doubting that we coordinate.
7:09
My worry is that common knowledge models have too many knife edge properties.
7:14
So whether or not there are timing frictions, whether or not there are differential interpretations of what's going on, whether or not there's an infinite number of messages or just an arbitrary arbitrarily large number of messages, all those can matter a lot in the model.
7:30
Yet actual coordination isn't that fragile.
7:32
So, isn't the common knowledge model a bad way to figure out how coordination comes about?
7:39
>> Well, it's so yes, so it could be there could be a direct link between a public conspicuous signal and the desire to coordinate.
7:47
And I think you're right, that's probably what happens most of the time.
7:50
But uh but in cases where there's doubt as to whether the signal has been received or how it's interpreted or where there's no way in which you can generate such a salient signal, that's when that's when we get what's often interesting in in life, namely you're trying to get inside someone else's head and you know, do they know what I know or do they know that I know it?
8:10
And a lot of our our not our routine automatic interactions but the ones where we kind of stop and think and calibrate our responses uh are are cases where recursive mentalizing does come into play.
8:25
Uh the choice of words in a fraught situation would be an example.
8:28
fraught situation would be an example. a uh a veiled bribe, a a sexual come on, a solicitation for a big favor like a donation where you put some mental effort into exactly how do I frame my overture uh so that they kind of catch my hearer catches my drift without
8:50
knowing that I know that they know which could change the basis of a relationship and that by the way I've I've skipped a step here which is that I suggest that human relationship ships like friendship, like authority, like romance, uh like uh bis transaction partners are all coordination games that are solved by common knowledge. Sometimes when we want not to change the
9:12
Sometimes when we want not to change the uh relationship in force, we actually take steps to avoid common knowledge uh as in inuendos and euphemisms where the message gets through but there's enough uncertainty as to whether your hero knows that that you know that they know that the relationship can survive the threat uh from the overture.
9:37
Doesn't differential knowledge play a big role in helping us understand what will happen?
9:41
So, say I'm part of an oligarchic elite and we all decide the actual lead dictator is about to fall.
9:49
That becomes common knowledge within the oligarchic elite. But that's not enough.
9:53
Someone has to move first.
9:53
Someone has to have differential knowledge that maybe they can take over or maybe they're the more courageous one or maybe they know who in the armed forces will support them.
10:03
And the actual model of cooperation is this optimal blend of differential and common rather than just common knowledge. No.
10:09
Yeah, I think that's I think that's right.
10:11
Uh often what once they make their move, they generate the common knowledge that allows everyone to coordinate.
10:17
But there has to be some kind of if either private knowledge or private initiative that uh that that that that launches the proliferation of common knowledge.
10:28
You I start off with the story the emperor's new clothes.
10:32
So why did that little boy uh blurt out that the emperor was naked?
10:37
Which then generated the common knowledge?
10:38
Now everyone knows that everyone knows. Why him? Why then?
10:40
I think part of the understanding of the story is that children uh being kind of naive and innocent about certain conventions and norms that govern adult life often do blurt out truths that adults don't.
10:56
They don't have the same internal uh kind of payoff structure that adults do.
11:00
They don't have the fear of ostracism or condemnation uh that adults do.
11:07
>> For a while now, we've had this thing called anonymous posters on the internet.
11:11
Do you think on average they make the world better or worse? Huh, interesting. Probably worse.
11:15
I haven't really thought. >> Okay.
11:19
But they help make some things common knowledge, right?
11:21
If there are things that are not common knowledge, some anonymous poster will say it.
11:26
There'll also be misinformation.
11:26
But is that a clue that maybe at the margin uh we have too much common knowledge rather than too little?
11:34
>> So it's a good question.
11:34
So I mean you asked the way you pose the question is is on average is it better or worse.
11:38
So it's kind of a matter of counting up the cases where it is a little boy saying the emperor was naked.
11:45
uh and and using I unlike the story using anonymity to be able to escape reprisals uh punishment for for um uh generating the common knowledge.
11:57
Uh you'd have to count up the number of cases where that happens compared to all of the misinformation, trolling, um hiding behind reprisals as a way of being uh aggressive.
12:09
You know, I think a lot of the shaming mobs, a lot of the trolling, uh, a lot of the degradation of discourse in social media comes because of that anonymity. Is it worth it?
12:23
Because sometimes it is a little boy saying the emperor is naked.
12:27
That that would be harder to tell.
12:28
But take even the cases where the anonymous posters are right.
12:34
So, as you and I know, there's a lot of alt-right views out there.
12:35
In general, I don't agree with them.
12:37
I think they're quite wrong.
12:39
But I also think that if some of them were correct, some of the conspiracy theories, it would actually be better if we don't all know them, that we would lose too much faith in elites.
12:48
And it's better that we do not have too much common knowledge again if those claims were true. Do you agree? Huh?
12:55
So you'd have to set I guess you'd have to set up the hypothetical.
12:58
That is are there cases where there's a noble lie where we're better off if everyone believes it?
13:04
believes it? I think a better a somewhat different case um other than an an outer lie would be a um a kind of um a gental benign hypocrisy where people where you have a common pretense and you're sometime collaborator Robin
13:24
Hansen writes about that in his book with Kevin Similar where he gives some examples and I actually repeat them and add some of my own of cases where it's not an outright lie but it's a case where you there's some things that are true that you avoid saying. So, uh Robin
13:37
So, uh Robin Johansson gives examples like uh if someone drinks from a whiskey bottle in a paper bag and on a park bench, uh you know, the paper bag isn't really fooling anyone.
13:51
Um but uh the police may choose not to arrest him under the pretense that he's not drinking an alcoholic beverage in public. Why would they do that?
14:02
Well, if they in tried to enforce the letter of the law, it would be too much like a police state.
14:07
It'd be a waste of resources.
14:08
On the other hand, if they failed to arrest someone who is overtly drinking, they'd be surrendering the uh authority of the rule of law in general.
14:18
And so the paper bag, what it does is it allows them to say the rule of law still holds in general even if we choose as we say and it's an interesting idiom to look the other way.
14:28
Now more conse and there are other cases in the in law enforcement that work like that. Uh escort services.
14:34
Does anyone really believe that someone's paying only for an escort uh but they get away with it?
14:40
Um uh uh uh to tobacco paraphernalia which is you know basically drug paraphernalia.
14:47
And of course there's it's a question of where you draw the line so that the law doesn't forfeit all of its authority.
14:54
By which I mean in going back to the top of the conversation, the common knowledge that the rule of law applies, which is necessary given that the police can't enforce everything 24/7.
15:03
Then on the international stage, things like um is Taiwan a sovereign state?
15:09
Uh well, of course it is, but the official policy of the United States is that it doesn't exist and of pretty much every other country as far as I know. Taiwan included, right? In a funny way. Taiwan included. Yes.
15:22
So, um I don't know if that would be an example.
15:25
It's not exactly a noble lie, but it's a common pretense.
15:29
And again, the reason you can spell out the reason, although if you spell out the reason, then that kind of undermines the reason.
15:35
But uh the reason of course being that it would be considered an affront, a challenge to China's sense of authority, prestige, standing, preeminence because it stakes a historical claim to all of his of a claim to all of historic China, which if it was explicitly breached by formally recognizing Taiwan, it would be a challenge to China's authority.
15:59
If nations act as if it was a sovereign state without saying it's a sovereign state, then China can maintain the the pretense that it is uh its claim is is valid.
16:11
Or yet another example is does Israel possess nuclear weapons.
16:16
Their official policy is, you know, we're not saying yes, we're not saying no.
16:19
Uh now, everyone knows that Israel has nuclear weapons.
16:22
If if there wasn't some suspicion, the weapons would be useless because they wouldn't be a deterrent.
16:27
On the other hand, if it was part of the official policy that Israel was a nuclear state, then its Arab rivals would feel that they had to acquire them as well so as not to lose face and standing or say determinism were true and free will were false.
16:42
It would be disastrous if we all believe that, right?
16:45
So, I think it's a good thing there's not more common knowledge because I think determinism is true. Okay. Yes.
16:51
Well, that's a whole other discussion.
16:54
I mean, one could say that determinism is necessary for the kind of um uh incentive structure that holding people respons responsible imposes.
17:06
Namely, if I condemn you, punish you, shame you, hold you responsible for bad choices, well, you're less likely to make those bad choices.
17:15
That's why we have a regime of holding people responsible.
17:18
So kind of paradoxically you need some degree of determinism for the idea of uh holding people responsible for their choices to have any effect.
17:29
Is a shelling point the same thing as common knowledge in your approach?
17:35
>> So a shelling point is uh I think it's more like common salience.
17:38
So it's it's closely related.
17:40
So just a shelling point would is a choice that people make to coordinate that has no intrinsic value other than the fact that everyone else knows that it exists.
17:50
It's something that that that that pops out.
17:54
Shelling's original example uh is a couple separated in New York.
17:57
This is before the era of cell phones.
17:59
How do they coordinate in the absence of common knowledge?
18:04
It can't just be that each tries to guess where the other one would go because you know if he you know if he knows that she likes to go to the bookstore so he goes to the bookstore.
18:13
Uh the problem might be that she knows he likes to go to the camera store so she might go to the camera store.
18:16
Well he knows that so he could maybe he should go to the camera store but then he knows that she knows that he knows that she likes to go to the bookstore so she might think oh he's going to go to the bookstore after all. You know add in an item.
18:28
an item. But if there's some point in the landscape that kind of commends itself as something that is salient enough uh obvious enough uh recommends itself uh like in Sheling's example he guessed that people would the couple would meet at the clock in the middle of Grand Central Station you know at noon
18:49
even if it wasn't where they would been separated simply because if you have to pick some place what place are you going to pick uh a shelling point is something that is salient enough to enough people that they ellight on it as a solution to a coordination problem in the absence of common knowledge, per se. >> So, Stephen, where would you go in New
19:06
>> So, Stephen, where would you go in New York to actually meet them?
19:09
>> Yeah, I know where you'd go to meet me, >> Grand Central Station.
19:13
>> Because it's it's it would pop into our minds and each of us knows it would pop into the other's mind because we both read Tom Shelling.
19:18
You know, nowadays, Grand Central Station isn't what it used to be and it's quite possible that it wouldn't be a shelling point.
19:24
wouldn't be a shelling point. that is a point that uh serves as a coordination uh equilibrium simply because each one knows that the other one knows that it will pop into the other one's mind and telling actually addresses that uh in he wrote in 1960 when Grand Central was much more influential and indeed I think
19:41
people did studies and they showed that people actually would as he predicted use the clock at Grand Central uh not as a plurality of choices uh but as Shelling notes because he he noted this is very fragile he said this is the point when it comes to actually choosing a what we now call a shelling point common salience. Uh this is where game
20:00
Uh this is where game theory, rational analysis, logic leave off and human whimsy, caprice, even poetry uh take over.
20:10
That is uh anything that pops into mind as a possible solution is a solution if it because it popped into your mind if it's likely to pop into the other person's mind.
20:22
So he said he he suggested, well maybe the couple will rely on whimsy and look for each other at the lost and found or maybe something came up in conversation the night before or maybe there was a guy with a sandwich board advertising a particular uh coffee shop.
20:36
Uh it uh when it comes to the points that people converge on because the other one is like knows that they're they will converge on it.
20:47
that there's no doesn't have to be any rhyme or reason other than human psychology. It pops into your mind. Okay.
20:54
So, where would you pick?
20:57
Essa Bagel somewhere, right?
20:57
Uh I I think it' be uh probably Time Square.
21:03
Maybe the statue of um what's his name?
21:05
Uh Sammy Khan, the uh there's a there's a kind of grandstand uh in Time Square with a statue of the the Broadway promoter.
21:14
Um that that might be >> I don't think you're going to meet many people. >> Maybe not.
21:21
>> Maybe the former World Trade Center site I thought of Moahan Station was not crazy.
21:27
>> No, the thing is it is and and Shelling points this out.
21:30
It is uh dependent on the context of that of your relationship with that person.
21:35
So it would be something like that may have come up in conversation, something that you remarked on, a quirk that one of you has that the other one knows about. Uh that's just it.
21:45
When it comes to a common knowledge generator, a U or a substitute for common knowledge really common salience, there is no predictability.
21:54
There needn't be any predictability from outsiders other than it pops into the likely to pop into the mind of the of the people involved.
22:06
Do you agree with Robert Alman's argument, at least as he makes it in the model, that if you meet an epistemic peer and you each announce your views, you ought to end up agreeing with them, even without debating the issue or exchanging information about the facts.
22:20
The mere fact that you have these two opinions, you should be able to converge. Yeah. So, if I I guess I do.
22:23
I mean, it is of course it it's it's uh dependent on the assumptions that go into the theorem.
22:31
This is the famous theorem that rational agents uh cannot agree to disagree or more accurately rational agents who have the same priors.
22:41
That is they come to a problem with the same understanding of the situation and hence the same degree of credence in uh a hypothesis before they even look at any evidence.
22:51
even look at any evidence. and they share their posteriors that that is each one kind of tells the other uh what their subjective estimate of the that idea is uh without without sharing the evidence that they use to come to that
23:09
assessment just what the assessment is then the theorem says those assessments have to be the same that is rational agents with the same priors and whose posteriors or common knowledge cannot agree to disagree and you know th those assumptions aren't you know are
23:26
seldom satisfied but if they're they are satisfied then then yeah uh and how does that actually manifest itself in some real world situation well uh I don't know if it was Alman himself it may have been I think geneopous who pointed out that a coralary of Alman's theorem which
23:42
sounds totally counterintuitive even even crazy is that people rational agents should not engage in speculative trade that is buying a stock in the expectation that it's uh uh that its val that you know that its value will rise and you can sell it at a profit. And
23:58
And this is exactly what all the investment analysts tell me.
24:02
Uh don't try to time the market.
24:05
Don't try to outguess other investors.
24:07
Just you buy an index fund and be content to share in the overall appreciation of the market.
24:12
That's a correlary of u Alman's uh agree to disagree theorem based on on the similar intuition.
24:20
Namely, since prices are common knowledge in at least in a in a large public market, if you have insider information, that that that changes the assumption.
24:32
But if since prices are common knowledge, any hint that you have that some company is coming out with an insanely great gadget or there's going to be a pent-up demand for this or that is almost certainly already reflected in the price.
24:45
So you can't do uh you can't reasonably expect to uh make a greater profit by uh speculating uh on private information.
24:57
>> Am I your epistemic peer when it comes to chocolate?
25:02
>> I've probably eaten a lot more of it than you have.
25:03
I've had it in more countries.
25:05
I've written a book on food.
25:07
So if I tell you chocolate is good and you shouldn't dislike it, >> should you just agree with me?
25:11
Will you in fact just agree with me? >> I I won't.
25:14
I'm one I'm one of these those freaks who just does not enjoy the taste of chocolate. >> Oh, that's fine.
25:19
But you can agree you should enjoy the taste of chocolate.
25:22
>> I ought to um >> you ought to. That's right. >> Yes, I ought to.
25:24
If I was if I had the same, you know, constitutional makeup uh as you then then perhaps I mean this is a case where the epistemic peers don't share priors.
25:34
And this could even when it came comes to pref I don't know if you could talk about the prior probability of a preference but when it comes to just the brute force gut feeling look I just don't like chocolate.
25:46
Uh you know that would be a case where it isn't irrational to not share those priors.
25:50
I mean that's just the way I'm that's the way I'm wired and and I I just don't like it as the >> I don't like mo I don't like most ballet but I admit I ought to.
25:58
I just don't have the time to learn enough to appreciate it.
26:02
>> Or take Alfred Hitchcock.
26:02
I would say North by Northwest while a fine film is really considerably below Rear Window and Vertigo.
26:09
Will you agree with me on that? >> Don't agree with me.
26:13
>> You think I'm not your epistemic peer on Hitchcock films?
26:15
>> Well, you know, preferences might are presumably different from beliefs.
26:17
That is >> no quality relative to constructed standards of the Canada, right?
26:24
>> Oh, well, they're okay.
26:24
So, that is the criteria exist external to both of us.
26:31
We both agree that they're reasonable.
26:34
Um uh so okay, we we agree that the criteria are reasonable.
26:40
We agree on everything that went on in the both movies.
26:43
We each share uh then we share our assessments.
26:49
I put Rear Window at a you know 7, you put it at uh point4.
26:52
Um is one of us irrational? Well, maybe we're not.
27:01
Okay, but you're going to budge now, right?
27:02
And you're going to agree that I'm right or we're not doing too well on this element thing, are we? We We aren't.
27:09
>> Cuz I'm going to insist North by Northwest, again, while a very good movie, is clearly below the other two.
27:15
Uh, you're going to insist. Yes. And I'm going to insist.
27:18
And I'm going to You might not agree with this, but I'm still convinced that if we had enough time, I could convince you.
27:24
And hearing that from me, you should exceed to the judgment.
27:28
>> Well, I I I should No, that that is right. And okay, I'm open.
27:30
Again, this is but crucially, we're not there can be no element of just raw subjective preference.
27:39
It has to be does it meet the criteria of, I don't know, the majority of Rotten Tomatoes or, you know, Roger Eert and and Pauline Kale and a panel of of of their peers.
27:50
So yeah, if it was if that existed then yeah, I'd be I'd be open and I'd say well I don't agree with you know Pauline Kale's criteria but you know who am I?
28:04
But you're not open on chocolate.
28:06
>> Again, you know, if if it's a matter of just do you like the taste there, it is there's, you know, dustabus uh non-estputandum.
28:16
When it comes to just preferences, I don't think that priors apply or or even meaningful.
28:21
But say I had that gene where, you know, cilantro to me tasted like soap. I don't.
28:26
But if I did, I would admit, well, it's really wonderful.
28:29
You know, I just have this disability.
28:32
and you have the chocolate disability.
28:35
>> Well, if you if you reframed it as a descriptive statement about other people's tastes, well, yeah, then it's just a question of asking them.
28:43
If it's a question of, you know, what is is is better.
28:47
Um, it better means uh, you know, I would pref I would choose it. there.
28:53
I I it seems to me that priors don't apply that that it's not a question of a descriptive belief, but there you really can have your own tastes.
29:01
Uh what's the best Bob Dylan album? >> Uh let's see.
29:08
I would say, you know, Highway 61.
29:13
Uh I think I'd say Bringing It All Back Home, Blood on the Tracks, and Highway 61 is number three.
29:18
>> What about Blonde on Blond?
29:18
What about Blonde on Blonde? >> Fourth. really okay.
29:23
>> It's a little too long.
29:23
Some of the songs are a little harsh in terms of the sound engineering I find.
29:28
Well, but it's very very good. Right. >> Yeah.
29:31
And you not not to mention that in all of these lyrics the songs there are some lyrics that are rather misogynistic by today's contemporary standards, but somehow he gets away with it even retroactively.
29:44
>> Is the liberal enlightenment gone forever? >> I sure hope not.
29:48
But uh but what do you think?
29:51
I mean, sort of betting odds.
29:51
It seems to me uh what we used to think of it is gone forever.
29:56
We may salvage some parts of it for the new thing that will be whatever that is.
30:01
I guess I'll say not gone forever, but certainly in retreat. Why not gone forever?
30:05
Um well, because it's got some built-in advantages.
30:11
Um, one of them is, um, reverence for for for science, which is, you know, the truth is what refuses to go away when you stop believing in it.
30:22
And so any method that is optimized for finding the truth has just that built-in advantage.
30:27
It can't it can be repressed. There can be censorship.
30:29
There can be irrationality.
30:31
But, you know, in general, something that finds the truth, it's got, you know, that going for it. and universalism.
30:37
Um, namely that if if liberal enlightenment depends on universal human rights, the the maximization of of uh human interests, the fact that we're, you know, ultimately uh kind of birds of a feather.
30:52
We all want to be alive rather than dead.
30:54
That there is something logically incoherent in saying I deserve rights cuz I'm me and you don't cuz you're you're not me.
31:00
you're you're not me. uh you know I can impose that by brute force but to the extent that we all have to agree on something we're not going to agree on you know me being imperial leader um so now those they aren't guaranteed to
31:16
succeed because there is brute force and and infirmities in human psychology and cognition and mob effects and all all the rest but it's nor is liberal enlightenment defense defenseless and the fact that some things tend to be, you know, I don't believe in historical momentum. I think that's kind of
31:34
I think that's kind of mystical.
31:35
But there are some things that they do look if not irreversible to have an awful lot of historical momentum.
31:40
So like human sacrifice, you know, all the ancient civilizations did it. Now it's kind of gone.
31:47
And I I don't think it's coming back. Why?
31:49
Well, a combination of we got enough knowledge about cause and effect to know that throwing a virgin into the volcano will not in fact bring a good weather or or make a a plague go away.
32:01
Um, some degree of of empathy for the average person so that we don't want to cut some the heart out of a beating heart out of someone's chest, you know, on average. Likewise, slavery.
32:14
It exists clandestinely, but we don't have any legal slavery anymore.
32:18
Democracy >> Moritania, don't we?
32:21
Don't we have more slaves in the world today than we did way back when by some counts?
32:27
>> Well, uh, as a not as a proportion of the world's population. >> Oh, sure.
32:30
But maybe it's the absolute number that matters.
32:33
>> Well, >> if you said to a slave, don't worry, you know, you're only onetenth of a thousandth of the world's population.
32:39
>> No, but >> wouldn't be that much of a constellation.
32:40
No, but if you said to a randomly selected person, um almost certainly you're not going to be a slave, then then that does matter.
32:46
Uh that is, you know, from behind a veil of ignorance, if uh an average person has some likelihood of being a slave and that's way down, then that matters.
32:56
But but in any case, legal slavery that is enforcable by the uh the state um and uh carried out in open slave markets no longer exists. Will it come back?
33:09
You know, I think probably not.
33:12
There are other things where it's not inexurable.
33:17
It's not monotonic, but it certainly seems to have a direction.
33:19
And that doesn't guarantee that it continues because there is no such thing as literal momentum.
33:25
literal momentum. But it does indicate that you know as there are greater means of exchanging information as there wider and wider pools of people exchanging uh information and having to agree on common interests there is some uh you know force some vector now again I don't
33:45
want to say it's going to win just that it exists and so it's not that liberal enlightenment is I think is gone forever it it and it isn't even gone I mean it's still more countries are dem democratic than not, even though if it's fewer than than it was 20 years ago. So, liberal
33:59
So, liberal democracy is kind of on its heels.
34:02
There is some retreat, but you know, we're still better off than we were in the, you know, the the '9s, the 80s, the 70s, the ' 60s, the ' 50s in terms of liberal democracy.
34:12
Um, so I don't think it's gone forever.
34:16
I think my view would be science will do fine, but even the enlightenment proper did not have universalism to begin with.
34:22
And the future for universalism to me looks quite grim.
34:26
And I think we will have a steady stream of low-level conflict across major powers.
34:33
So China does all these cyber attacks on us.
34:35
And they would have been acts of war way back when and we just put up with it.
34:39
Russia sabotages EU infrastructure.
34:41
That might have been an act of war way back when. We just put up with it.
34:46
And we'll just be creeping up up up that threshold.
34:48
not to nuclear war, but we'll get as close as we can get without deterrence really having to kick in and the world will just be more violent and have more conflict. Why is that wrong?
34:59
>> I I think that's that's interesting because as you say, it doesn't result in war and war would be much worse.
35:04
So the invasion of Ukraine, you know, rewinding the clock 100 years, that could have been like the uh, you know, assassination of the arch duke in Sievo.
35:15
It could have been that NATO decided to go allin and uh something like World War I or even worse could have happened.
35:20
As awful as Ukraine is, and that is awful, it's not as bad as World War I.
35:25
And uh if the these cyber attacks, these insults because of a norm threatened by Putin, no doubt, and and by by she that you don't invade your neighbors, you don't try to rearrange borders by force.
35:40
borders by force. um we we haven't had another world war and um you know if what you said is correct you know things could suck uh but still they could suck much worse if we had a world war and I think it is some less likely than it was 100 years ago if you had to model in as
35:58
fundamental and abstract a set of terms as possible not relying on the names of individuals very much why liberal enlightenment is in retreat today what does your model look like uh I I think the basic answer is that liberal enlightenment is not particularly intuitive. It's uh it doesn't really fit
36:14
It's uh it doesn't really fit our uh human nature in the sense of our conception of how um our our societies ought to be organized.
36:24
What's much more natural is tribalism uh deference to authority um uh conformity.
36:32
Uh I think liberal enlightenment is a great idea.
36:35
It's you like a lot of counterintuitive discoveries of organized rationality.
36:40
It is commendable even if not intuitive.
36:44
So I think there's a natural backsliding unless it's constantly propped up.
36:53
People are going to say well we need a well the solution to our problem is we need a strong leader.
36:56
Uh what is a strong leader?
36:58
Well we we once he's a leader he gets to say what's what's what.
37:02
The idea that a leader just temporarily presides and is kind of bound by fiduciary duties laid out with, you know, checks and balances and safeguards, it's a fabulous idea.
37:12
It's not particularly intuitive in the same way that markets are not particularly intuitive.
37:18
Uh in the same way that science is not particularly intuitive, but they're still good things.
37:21
but they're still good things. So it's not it's it's kind of natural uh to to think why do people have u superstition belief in paranormal belief in woowoo why do they uh not grasp certain you know economic fundamentals like you know
37:40
uh trade is good is not it's it's not zero sum why is you know congestion pricing in Uber rational even though people hate it there probably a lot of cases where people just left to their own devices kind of backslide away from liberal enlightenment principles. >> Do you think we have stirring liberal
37:56
>> Do you think we have stirring liberal thinkers of today who are younger than 55 years old? Not stirring enough.
38:04
I think we do have such such thinkers.
38:06
Um, you know, uh, oh, uh, you know, Derek Thompson, uh, Martin Budri, um, uh, Kathy Young, although I guess I don't I don't know what I haven't kept track of her age.
38:21
Uh, she's she's younger than 70. Uh, good.
38:26
But I, yeah, there are very young people.
38:29
Um, uh, I don't know, but you know, Ezra Klene, I'm talking now about kind of journalists.
38:36
If you go into academia, there would probably be even more.
38:38
And there are other people that I think our audience may not have heard of that are in my uh you know, email correspondents.
38:48
Um are they stirring enough?
38:48
Are they like a whatever Savoy Xek has?
38:51
You know, do they have it? You know, probably not.
38:58
Although that's might be an inherent problem in that influence by charisma kind of goes against the whole idea of liberal enlightenment where it should it ought to just be the the the the quality of the content of your ideas not who looks more dashing in a scarf.
39:14
Uh so little enlightenment it kind of is at a disadvantage in that regard and that it it oughtn't to rely on charismatic thought leaders.
39:27
How much of a liberal was Napoleon?
39:29
>> Not well wait, let me back.
39:31
>> Wasn't he quite a liberal?
39:31
He just loved war too much and he was an egoomaniac and he wanted to take over other countries and that killed millions.
39:37
But wasn't a lot of his program quite liberal? >> No, you're right.
39:41
There there was there was certain, you know, kind of well rethinking institutions and laws uh introducing rational things like the metric system.
39:51
um you know uh but but yeah it could be that his own uh malignant nar narcissism uh subverted his the liberal parts of him.
40:05
>> Doesn't everyone at Harvard deserve an A minus or an A?
40:09
>> They made it into Harvard, right? >> Okay. >> Same with Stanford. >> Yes.
40:12
So, I actually have data on this because that's what our dean told us, not with a whole lot of conviction, when we had a meeting about grade inflation, and she said, "Well, we like to think of it as grade compression.
40:23
Uh, maybe students are just getting better and better and better."
40:27
And I I just know that that's not true because I have records from performance on my multiplechoice test in my intro psych course, which I've taught for more than 20 years.
40:37
And uh, so it's pretty constant yeartoear.
40:40
and there's no subjective sliding uh grade and the the performance has gone down.
40:46
Uh so students are not doing as well as as they used to.
40:50
Um and I've been forced just so so as not to drive away all my students to inflate my grades.
40:55
I give way more aids than I used to.
40:57
Uh just so I don't have an empty lecture hall and uh it's not because the students are getting better and better.
41:04
And we know that a lot of Harvard students are admitted on non-marritocratic uh criteria.
41:08
Even putting aside, you know, forget about, you know, the the whole uh hot button of of racial preferences, but legacies, athletics, um social action, they sorted clothes for the homeless, they circulated a petition, they hung out with artists in an artist colony.
41:25
Uh uh Harvard does not admit students on on meritocratic grounds.
41:33
>> So, if you're the boss, what would the median grade at Harvard be? >> The median grade? Uh >> median grade? Yeah. >> Let's see. Um, B minus.
41:45
>> And how many students at Harvard, what percent do you think essentially deserve A minus or A? They're just very smart. They work hard. They're great.
41:52
Maybe Harvard classes are harder, but they just ought to have good grades because they're wonderful. >> A quarter. 25%. >> Only a quarter.
42:02
I'd be shocked if it's only a quarter.
42:04
Your admissions department must be terrible then. >> They It is terrible.
42:06
Well, I've argued that as much.
42:08
They use holistic criteria, which is mumbo jumbo.
42:10
It's it's it's mysticism.
42:12
Uh the the reason I say that is, you know, my I I think I'm a pretty good lecturer.
42:19
And I mean, for whatever that's worth, everyone thinks they're a good lecturer.
42:23
>> No, I don't doubt that. I'm sure you are.
42:24
>> But my my my lecture all is, you know, two weeks into the semester or 3 weeks into it's half empty.
42:28
So, anyone whose Fanny is not in that seat, I think, does not deserve an A cuz, you know, it's not asking a lot.
42:36
It's like less than 3 hours a week.
42:38
And they get to hear, you know, me.
42:41
I mean, I don't think I have a big eagle, but people in the real world actually pay money for tickets to my events.
42:46
And for Harvard students, it's already paid.
42:49
They still stay away because they have more important things to do like pulling an ore or singing in an ac capella group. Do they deserve a name?
42:56
I frankly think they don't.
42:59
Are you discouraged that linguistics has had so little impact on current large language models? Yeah, I am.
43:04
And um you know I don't know whether it's that linguists themselves have not u made themselves useful enough have been part of the conversation or I mean the cynical view is linguistics has just been on the wrong track and it has nothing to contribute.
43:21
Um I I tend to think more the former.
43:24
Now, there are some features in large language models that actually implicitly build in some of the insight of linguistics.
43:31
The fact that you've got all of these hidden layers uh is an acknowledgment that the regularities in language and the regularities in thought that language expresses are um abstractions.
43:41
They're not uh triggered by individual words, but they're um patterns and patterns of patterns, groupings, and larger and larger groupings.
43:55
the fact that the one of the breakthroughs in the this generation of large language models is the uh what I call attention attention heads which are basically processors that look at uh temporally separated words uh in search of regularities as opposed to adjacent ones.
44:14
The fact that they're not trained on words themselves, but on um tokens that correspond to more or less to morphs.
44:24
The fact that they're trained on embeddings, which is not again not the words themselves, but some backhanded indirect representation of meaning, in that case, the con context that a word typically occurs in.
44:36
So there are a lot of discoveries of linguistics that are implicitly built in.
44:42
But linguistic theory in the sense of what you see in a a journal like linguistic inquiry aren't don't have direct counterparts.
44:50
Now maybe they wouldn't need all of those gigawatts and all of those uh GPUs if they uh kind of were able to cut to the chase quicker by building in linguistic insights.
45:01
My hunch is yes, but I the models are so complex that I can't say for sure.
45:09
>> Does it favor Chomsky?
45:09
Does it favor the view that language is simpler than we thought?
45:14
>> Where does it push your intuitions? >> Yeah.
45:16
Um the it has changed my intuitions to that these models are capable of more than I would have guessed that they're capable of given how they're built.
45:26
But that may show that my intuitions break down when it comes to billions, tens of billions, hundreds of billions of parameters.
45:37
uh uh 90 hidden layers training that would be the equivalent of 30,000 years if a a child was was listening to it.
45:45
listening to it. my intuitions may be useless of what you can accomplish uh at those scales which le leads to a kind of sort of ignorance of um I think all of us so far until we probe the models to find out what's in the black box and what's
46:03
responsible for their intelligence does it mean that the linguists and cognitive psychologists were wrong all along uh in saying that you need rules algorithms proposition conditions, mental models, that it's all just massive statistical patterns. Or is the discrepancy between
46:21
Or is the discrepancy between the fact you have to train these models for the equivalent of 30,000 years and a human child can speak in 2 years say that it's almost a coincidence that they're as capable as they are.
46:33
they're kind of capitalizing on all this input that we humans have put out there and that they're doing something humanlike in a very different manner than than the human brain.
46:45
I I tend to think the latter and the fact that they are so vulnerable to confabulation, hallucination, blends, as my former student Gary Marcus uh insistently points out is a kind of hint that they're doing things in a different way than the human brain is doing things.
47:04
How long do you think it will be before they can write a poem say as good as the median poem by Pablo Nuda?
47:12
>> The thing is that the fact that Pablo Nuda didn't write it it itself goes into our assessment of how good a poem it is that human work >> a blind taste test, right? >> Yeah. Blind taste test.
47:21
Um yeah, I don't I don't know.
47:26
They're already better at doing things probably like composing limmericks or higgledy piggledies on the fly.
47:31
uh they're not, you know, great literary quality, but they are often very stringent formal criteria.
47:38
Uh they're remarkably good at that.
47:41
Um the thing is that in again we're going back to the uh you know uh north by northwest versus rear window.
47:52
Um is the qu what goes into assessment of quality and to what extent is there such a thing as objective quality of of works of art?
47:59
You know, I tend to think that our assessment of a work of art isn't completely extricable from our knowledge of its providence.
48:10
Which is why a fake Da Vinci, an imitation Da Vinci, uh is just, you know, not the same as a real Da Vinci in terms of the experience, in terms of any judgment.
48:20
Uh how it's produced goes into our assessment of what kind of work of art it is.
48:26
Before my last two questions, let me just plug your book again.
48:29
It is called When Everyone Knows That Everyone Knows. Dot dot dot. The dots are important.
48:36
And the subtitle is Common Knowledge and the Mysteries of Power, Money, and Everyday Life. A fascinating read. Last two questions.
48:43
First, what's your favorite song on Rubber Soul?
48:49
>> Uh, mine is You Won't See Me.
48:49
I'll announce in advance, but I know >> I really like that song.
48:53
I think I've just seen a face. >> It's a good pick.
48:56
I like the sort of bluegrass rhythms behind it all. >> Yes.
49:00
And just the kind of insistent u uh you know rhythm and the piling up of of line on line. Final question. What will you do next?
49:12
>> Oh uh one possibility is to revisit the uh all the indicators of progress.
49:20
What's the report card on the world 10 years after I published um enlightenment now?
49:24
uh what progress has continued where have we backtracked and it in a sense answer your question what are the prospects for liberal enlightenment Steven Pinker thank you very much thanks so much Tyler always a pleasure [Music] [Music]