Speaker critiques Goldman Sachs' game theory interview question as logically flawed and impractical
Summary
The speaker argues that the standard backward induction solution to the Goldman Sachs lion puzzle is theoretically fragile and practically dubious. He compares the problem to the finitely repeated prisoner's dilemma, suggesting that strict rationality assumptions fail when players can deviate or learn. The speaker contends that higher-order rationality becomes unstable as the number of lions increases, introducing doubt into the equilibrium. He further illustrates the flaw by referencing the Paradox of the Unexpected Hanging, showing how backward induction can lead to incorrect real-world outcomes.
Key takeaways
- 1 The speaker argues that if a lion deviates from the 'safe' strategy, it introduces doubt into other lions' beliefs, making the 'perfectly rational' equilibrium unstable.
Unidentified speaker (speaker_0) · 26:08
Episode sections
- 0:00 – 2:09
Introduction and Game Theory Setup
Speaker introduces a viral Goldman Sachs interview question involving 253 lions and a poisoned steak, arguing that the standard game theory solution is flawed. He notes the question has gained traction on social media and sets out to critique the accepted answer.
- 2:09 – 17:14
The Standard Solution and Nash Equilibrium
Speaker details the 'correct' answer: using backward induction, lions will eat the steak if the number is odd (253) and refrain if even. He explains this forms a subgame perfect Nash equilibrium based on common knowledge of rationality.
- 17:14 – 25:11
Critique via Finite Repeated Prisoner's Dilemma
Speaker compares the lion problem to the finitely repeated prisoner's dilemma, arguing that backward induction logic fails in practice because it ignores the possibility of learning or deviation from strict rationality in finite games.
- 25:11 – 28:25
Challenging Rationality Assumptions
Speaker argues that if a lion deviates from the 'safe' strategy (eating the steak when 252 remain), it introduces doubt into the other lions' beliefs, making the 'perfectly rational' equilibrium unstable and potentially irrational to maintain.
- 28:25 – 31:17
Lion Paradox and Rationality Assumptions
Speaker_0 uses a lion and steak analogy to critique game theory's backward induction, arguing that observing a violation of the 'even number' rule undermines the logical chain. He contends that rationality cannot be assumed to hold perfectly across multiple layers of strategic reasoning.
- 31:17 – 32:38
Finite vs. Infinite Games and Discount Rates
The speaker explains that the 'final round' problem in finite games can be mimicked in infinite games through discount rates or random coin flips determining continuation. He argues that without a known final round, backward induction logic fails because players cannot identify the last move.
- 32:38 – 34:00
Game Theorists vs. Street Players Payoffs
Speaker_0 presents a hypothetical scenario where game theorists playing a finite prisoner's dilemma earn lower payoffs than 'street people' who cooperate for most of the game. He uses this to argue that claiming game theorists are 'more rational' is invalid if their strategy yields worse outcomes.
- 34:00 – 37:15
Complexity of Multi-Layer Rationality
The speaker clarifies that while simple rationality (eating when safe) is intuitive, higher-order rationality requires assuming others are reasoning identically. He argues this chain of 'I know that you know that I know' becomes fragile and dubious as the number of lions increases.
- 37:15 – 39:42
Paradox of the Unexpected Hanging
Speaker_0 introduces the 'Paradox of the Unexpected Hanging' from the 1940s/50s, linking it to the lion problem. He describes a judge's decree that a prisoner will be hanged on a weekday when he is surprised, which the prisoner's lawyer argues logically prevents the hanging.
- 39:42 – 42:21
Resolution of the Hanging Paradox
The speaker recounts how the lawyer's backward induction eliminates every day as a possible hanging date, leading the prisoner to feel safe. However, when the guard arrives on Wednesday, the prisoner is surprised, proving the judge's statement true despite the logical deduction.
Claims worth verifying
The speaker references a Goldman Sachs interview question involving game theory that has circulated on social media.
Unidentified speaker (UNKNOWN) ·28:25
How to verify: Search social media for 'Goldman Sachs interview question game theory lions' and cross-reference with the episode title 'Why Goldman's Game Theoretic Job Question Is Dubious'. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.
The speaker claims that in the lion paradox, if 252 lions are present, the 252nd lion eating the steak is not necessarily a mistake because it breaks the 'even number' rule illusion.
Unidentified speaker (speaker_0) ·30:21
How to verify: Review the transcript at 30:21-30:34 where the speaker argues that breaking the spell of the even-number rule allows for eating to occur, challenging the standard backward induction solution. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.
The speaker states that a finite repeated prisoner's dilemma can have the same strategic flavor as an infinitely repeated game if a coin flip determines continuation.
Unidentified speaker (speaker_0) ·31:17
How to verify: Review the transcript at 31:17-31:31 where the speaker explains using a coin flip to decide if another round is played, creating uncertainty about the final round. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.
The speaker asserts that game theorists playing a finite prisoner's dilemma for 1,000 rounds would defect the whole time, while street people would cooperate for 850 rounds.
Unidentified speaker (UNKNOWN) ·32:38
How to verify: Review the transcript at 32:38-33:07 where the speaker outlines the hypothetical payoff comparison between game theorists and 'two Joes from the street'. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.
The speaker claims that the 'Paradox of the Unexpected Hanging' was a major topic of argument among mathematicians and logicians in the 1940s or 50s.
Unidentified speaker (UNKNOWN) ·37:40
How to verify: Search academic databases or historical records for 'Paradox of the Unexpected Hanging' and verify its prominence in literature during the 1940s-1950s. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.
The speaker describes the judge's decree in the hanging paradox as occurring on a Sunday morning, with the hanging to happen at 9 a.m. on a weekday.
Unidentified speaker (speaker_0) ·38:15
How to verify: Review the transcript at 38:15-38:27 where the speaker sets the scene of the judge's announcement regarding the timing of the hanging. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.
The speaker states that the lawyer's logic eliminates Saturday as a possible hanging day because the prisoner would not be surprised.
Unidentified speaker (speaker_0) ·39:17
How to verify: Review the transcript at 39:17-39:42 where the speaker explains the backward induction step ruling out Saturday due to the lack of surprise. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.
The speaker claims that in the resolution of the paradox, the guard arrives on Wednesday to hang the prisoner, surprising him.
Unidentified speaker (speaker_0) ·41:31
How to verify: Review the transcript at 41:31-41:39 where the speaker concludes the story with the Wednesday hanging and the prisoner's shock. For recent events, verify 2026-08-14 through 2026-11-14. Use another period only when the speaker explicitly names it.
Do your own research
- Verify the original Goldman Sachs interview question and its accepted solution via reputable financial news archives or the firm's official communications.
- Review academic literature on the 'Finitely Repeated Prisoner's Dilemma' to compare theoretical predictions with experimental data on human behavior.
- Research the 'Paradox of the Unexpected Hanging' (also known as the Surprise Examination Paradox) to understand its logical structure and proposed resolutions.
- Check primary sources on 'backward induction' and 'subgame perfect Nash equilibrium' to assess the mathematical validity of the speaker's critique.
- Investigate empirical studies on 'higher-order rationality' and 'common knowledge' in game theory to see if the speaker's claims about instability are supported by research.
- Look for counter-arguments from professional game theorists who defend the standard solution to the lion puzzle.
- Consider the difference between theoretical models of rationality and observed behavior in real-world strategic environments.
Full transcriptexpand
- 0:00
- Unidentified speaker (speaker_0): InFi, episode 162. It's possible that when there's 251, even though using the backwards induction logic, you would think I'm a dead, I'm a goner for sure. Maybe not. Because of again, the fact that we've now broken the spell and shown you can't conclude that if there's an
- 0:20
- Unidentified speaker (speaker_0): even number, nobody's eaten eating it because I just did it. I was a 250-second 252nd line and I ate that steak. Welcome to InFi. the future of finance.
- 0:31
- Unidentified speaker (UNKNOWN): Hey
- 0:32
- Unidentified speaker (speaker_0): everybody, welcome back to the InFi podcast. This is going to be an unusual episode. I am responding to something that has been blown up on social media.
- 0:42
- Unidentified speaker (UNKNOWN): This
- 0:42
- Unidentified speaker (speaker_0): guy, Mercurious, posted what is ostensibly an interview question that Goldman Sachs puts out. And it involves
- 0:52
- Unidentified speaker (speaker_0): game theory. And so it was posted I I saw it on X,
- 0:57
- Unidentified speaker (speaker_0): the website formerly known as Twitter. And like I say, it was it was blowing up. It's like the the classic that midwit meme that goes around where the lay person reading this is we'll get to it in a second, folks, of course, thinks, oh that's a silly question. And then there's what the smart person would think IQ of
- 1:17
- Unidentified speaker (speaker_0): 110 thinks is the right answer. And then I'm gonna come in on this episode of InFi and say, No, no, no. Actually, I don't think that
- 1:27
- Unidentified speaker (UNKNOWN): the
- 1:27
- Unidentified speaker (speaker_0): uh highbrow people lecturing everybody on what the obviously correct answer is. I don't think they're right. And I think it's instructive to go through this. This isn't really just something that was making the rounds. I think it's relevant. And also, I was thinking, geez, is this really appropriate for infy, even though I think it's interesting and I think you folks will find it interesting. Again, this was a Goldman Sachs interview. Goldman Sachs,
- 1:47
- Unidentified speaker (speaker_0): clearly part of the financial sector. So there you go. It's definitely relevant.
- 1:51
- Unidentified speaker (UNKNOWN): So
- 1:51
- Unidentified speaker (speaker_0): why don't we jump right in here? Okay. So this is the tweet in and question. And in case you are not able to read it yourself because you're at the gym or something, let me go ahead and read this out loud. So just to give credit, was from this account, Mercurious.
- 2:09
- Unidentified speaker (UNKNOWN): And
- 2:09
- Unidentified speaker (speaker_0): he says, How would you answer this common Goldman Sachs interview question? So here's the question. There are 253 lions and one steak. The steak is poisoned. If a lion eats it, the lion falls asleep. While asleep, the other lions can eat that lion. and then fall asleep themselves.
- 2:29
- Unidentified speaker (speaker_0): All lions are perfectly rational and want to survive. Will any lion eat the steak? Okay.
- 2:36
- Unidentified speaker (speaker_0): So
- 2:37
- Unidentified speaker (speaker_0): there were a lot of people that were raising obvious types of objections, like
- 2:44
- Unidentified speaker (UNKNOWN): we
- 2:45
- Unidentified speaker (speaker_0): talking about 253 lions all in one spot. It doesn't make any sense, and they'd be running away. and or you know whatever fighting over them between themselves or they'd go find gazelle they're not gonna eat. What do you mean it's poisoned if it just puts them to sleep? What does that mean? Okay. What the lions are gonna be cannibals now? Okay, lots of stuff like that. All perfectly
- 3:04
- Unidentified speaker (speaker_0): valid, but clearly you gotta play by the rules of the game. Okay. This is obviously the point of this is not an exercise in national geographics. adventures or something. Like this isn't supposed to give you tips on how to survive in the wild. This is obviously just a brain teaser. And the reason Goldman Sachs apparently
- 3:24
- Unidentified speaker (speaker_0): asked this of some job candidate is that they wanted to see how would you reason through this. Okay. So things that are technically missing from this that if it were in a game theory series class somewhere that they probably would have supplemented the question with, as it stands, it doesn't say why
- 3:44
- Unidentified speaker (speaker_0): would the lions eat the steak at all, right? So besides just when it says all lions are perfectly rational and want to survive, they must also mean that yes, the lion doesn't want to be eaten, but it also gets some benefit from eating the steak, and that if there's a lion who now is out because he
- 4:04
- Unidentified speaker (speaker_0): either ate the original steak or ate another line who was in that chain, that the benefit, if you're the next line in line, no pun intended there, other things equal, you'd want to eat the now asleep lion that that would help you to get a meal. All right. So they didn't technically say that here, but we can give the benefit of the doubt and say, okay,
- 4:24
- Unidentified speaker (speaker_0): yeah, that's in there too. So now that I've just done some of the housekeeping here, let me first walk you through what clearly they think the correct answer is, and that what a lot of the people on X were saying, Oh, here's the answer. I can't believe so few people got it. Okay, so it's easiest to start with
- 4:44
- Unidentified speaker (speaker_0): a smaller number. Let's just say there's one lion, right? And there's a steak that's sitting there,
- 4:49
- Unidentified speaker (UNKNOWN): and the
- 4:49
- Unidentified speaker (speaker_0): lion can eat the steak. but the downside is, oh, it'll knock you out for a little bit. And you'd be vulnerable if another lion were there who could then in turn eat you if the lion so chose. But if there's only one lion, then that's not an issue. So there's no downside. If there's one lion there, he will clearly eat the steak taking that.
- 5:09
- Unidentified speaker (speaker_0): Right? Okay. What if there's two lions? Well, you might think oh, by the way, another quibble people had was to say the question is clearly envisioning that like each line is in sequence and that it would make as opposed to like five lions just all trying to eat the steak
- 5:29
- Unidentified speaker (speaker_0): simultaneously, or even more plausibly, suppose some lion does eat the steak and then passes out. Why would you just assume? only one lion out of the remainder would be the one to eat that line and then pass out like sequentially? But as we'll see, that's clearly what the question wants you to do when you
- 5:49
- Unidentified speaker (speaker_0): give the answer is to assume it works like that. That it's almost like the lions are in one big queue, and that the one at the front of the line gets to decide, am I gonna eat this or not? And his decision kind of holds everybody else up. Right. So that's that's the logic that they have in mind here in terms
- 6:09
- Unidentified speaker (speaker_0): of the game flow. So again, some people quibbled and said, Well, they didn't spell that out. Why would you assume that? But okay. So back to my walking through what I think they clearly intend the correct answer to be. If there's two lions, imagine there's like with the first lion that's sitting there looking at the steak and then a lion. right behind him,
- 6:28
- Unidentified speaker (UNKNOWN): or
- 6:28
- Unidentified speaker (speaker_0): I guess they could be circling it. And the issue is, would either of them be the first mover to eat that steak? And
- 6:37
- Unidentified speaker (UNKNOWN): it
- 6:38
- Unidentified speaker (speaker_0): seems pretty solid to say no, not with the assumptions given, because
- 6:43
- Unidentified speaker (speaker_0): if one lion Line eats it, then he's out, and then there's only one other line remaining. And the question is: what would that line do if he found himself in that scenario? And he would eat that passed out line for all the same reasons that we walked through as to why, if there's just one line there, he'll eat the steak.
- 7:03
- Unidentified speaker (speaker_0): Right? There's no downside. If there's only two lines, the first line happens to eat the steak and is out.
- 7:08
- Unidentified speaker (UNKNOWN): The
- 7:08
- Unidentified speaker (speaker_0): a line, the line The lion who's still awake, has no downside to eat that line, lion, so he would. But now if you're with me so far, and that seems like it's pretty obvious, like you don't have to assume anything except that
- 7:21
- Unidentified speaker (UNKNOWN): a
- 7:21
- Unidentified speaker (speaker_0): hungry lion who's capable of realizing there are no downsides here is gonna eat the meal that's in front of him.
- 7:28
- Unidentified speaker (UNKNOWN): Well
- 7:28
- Unidentified speaker (speaker_0): then that's why now,
- 7:30
- Unidentified speaker (UNKNOWN): if
- 7:30
- Unidentified speaker (speaker_0): there's just two lines originally and they're both rational, and it says perfectly rational was the actual phrase they used in the question.
- 7:39
- Unidentified speaker (speaker_0): What they want you to think is, oh, each lion being perfectly rational is going to assess the situation and say, I wouldn't eat the steak or I shouldn't eat the steak, because if I'm under, then the sole remaining lion is clearly going to eat me. And so, since of the two lions, they both
- 7:59
- Unidentified speaker (speaker_0): know that they each refrain from eating the steak, stake, and that's it. Okay. Don't worry, folks, I'm not going to go all the way through 253 of these, but I just want you to see the pattern.
- 8:08
- Unidentified speaker (UNKNOWN): So
- 8:08
- Unidentified speaker (speaker_0): now suppose there's three lions. And so here the argument goes each lion is perfectly rational. They can walk through the logic that we just entertained ourselves. And so the lion, each lion is going gonna to realize, oh,
- 8:25
- Unidentified speaker (UNKNOWN): if
- 8:26
- Unidentified speaker (speaker_0): I had get to the stake first and I'm the, you know, I I go ahead and eat it before the other two have a chance, what's gonna happen? I'm gonna take a nap, and then those two lions, while I'm out, are gonna be in the same strategic position as two lines looking at the stake would be. And we've just
- 8:45
- Unidentified speaker (speaker_0): concluded
- 8:46
- Unidentified speaker (UNKNOWN): that
- 8:46
- Unidentified speaker (speaker_0): if the lines are perfectly rational,
- 8:49
- Unidentified speaker (UNKNOWN): if
- 8:49
- Unidentified speaker (speaker_0): there's ever a scenario where there's two lines. looking at the steak, neither of them's gonna eat it. So now when there's three lines originally, they all know, oh, if I could be the first one there and gobble that thing down, yes, I'm gonna pass out, but I'm gonna be perfectly safe. Because again, that's the same strategic
- 9:09
- Unidentified speaker (speaker_0): situation. And since I know these other two lines are rational, perfectly rational, in fact, they're thinking the same way that I am. And so therefore, I'm totally safe. So incidentally, that's another assumption that the question didn't explicitly state, and that some people, especially those who had studied game theory, were pointing out
- 9:29
- Unidentified speaker (speaker_0): that it's not enough to make this solution go through that I'm walking you through here. It's not enough that all the lines are perfectly rational, even conceding or stipulating. what, quote, perfectly rational means in this context and what the Goldman Sachs people mean by that term. Even if we grant all that, the way this
- 9:49
- Unidentified speaker (speaker_0): ostensible solution is going to go through is each line not only is perfectly rational, but knows that the other lines are perfectly rational. And you might think, oh, that's you can stop there, right? No. They also have to know. that all the other lines know that they're all perfectly rational and that
- 10:07
- Unidentified speaker (speaker_0): and they have to know that they know that they know for at least in this case up through the two hundred and fifty third level of that. So in game theory, what they do to get that across is they'll say it's common knowledge that the lines are perfectly rational in this kind of a setting. To mean like that's
- 10:27
- Unidentified speaker (speaker_0): shorthand to say, oh, I don't know. The fact that the sun is hot is common knowledge. Right. So it's not just that everybody knows it. Everybody knows that everybody knows it. And everybody knows that everybody knows that everybody knows it and da-da-da-da-da, as far as you want to push it. That's what it means to say something is a certain belief is common
- 10:47
- Unidentified speaker (speaker_0): knowledge when you model the this stuff formally. Okay. So I'm just saying technically you need that to be true for this demonstration to go through that I'm walking you through here. I'm sure just about all of you have already seen it, but let me just go ahead and do it a little bit more. Okay, so again, if there's three lions, then the argument is
- 11:07
- Unidentified speaker (speaker_0): whichever one is able to get to the stake steak first can go ahead and eat it safely. Because knowing
- 11:13
- Unidentified speaker (UNKNOWN): if
- 11:13
- Unidentified speaker (speaker_0): there's just two looking at me. taking a nap, neither of them is gonna go after me. Okay. What if there's four lions? Well, now they're gonna say, well, if I eat the steak, I'm out. And then at that point, there'd be three lions looking at me. And we've just concluded
- 11:32
- Unidentified speaker (UNKNOWN): if
- 11:32
- Unidentified speaker (speaker_0): there are three lions lines looking at the steak, whoever gets to it first is gonna safely eat it.
- 11:37
- Unidentified speaker (UNKNOWN): And
- 11:37
- Unidentified speaker (speaker_0): so if there's four lions, lines, they all going through that logic realize, no, I don't want to eat that steak because then there'd be three lions looking at me taking a nap, and we all know the first one to get to me would chomp me. Right? Okay. So look at what the pattern is so far. If there's one lion looking at a steak, he's gonna eat it. If
- 11:57
- Unidentified speaker (speaker_0): there's two lions lines looking at the steak, neither of them will eat it. If there's three lions lines looking at the steak, one of them will eat it. If there's four lions looking at the stake, nobody will eat it. And you can just keep doing that. And the pattern is if you buy the argument and the logic I was just using, the pattern that emerges is if there's an even number of lions, nobody eats
- 12:17
- Unidentified speaker (speaker_0): the steak. If there's an odd number, the first lion eats And the steak because he and then that's it. He takes a nap, and every all the other lines are just staying. staring at him until he wakes up again. So notice how that's all consistent.
- 12:31
- Unidentified speaker (UNKNOWN): You
- 12:31
- Unidentified speaker (speaker_0): could say that forms a Nash equilibrium. And so that even an odd, that that's how it works, in case I didn't finish the train of thought. All right. And so that's why, by the way, just said go ahead and make sure we I give you the interim punchline. There's going to be a bigger punchline later as I give you my thoughts, but
- 12:45
- Unidentified speaker (UNKNOWN): on
- 12:46
- Unidentified speaker (speaker_0): this train of thought for what presumably the official correct answer is. Because that was a very specific thing that they said there are 253 lions. Then just say there's a group. So clearly the number matters. Otherwise, why would they give you that information? And so the reason it matters is once you build up this pattern, you say, Oh, because there's
- 13:05
- Unidentified speaker (speaker_0): 253,
- 13:06
- Unidentified speaker (UNKNOWN): and
- 13:07
- Unidentified speaker (speaker_0): that's an odd number, then
- 13:10
- Unidentified speaker (UNKNOWN): one
- 13:10
- Unidentified speaker (speaker_0): lion, whoever gets that stick steak first, will eat it,
- 13:13
- Unidentified speaker (UNKNOWN): and then
- 13:13
- Unidentified speaker (speaker_0): the other 252 lions are going to sit there looking at them, taking a nap, and that's
- 13:17
- Unidentified speaker (UNKNOWN): it. So
- 13:18
- Unidentified speaker (speaker_0): I'm saying notice just to flesh that out, that if you stated that, that would form what's called a Nash equilibrium.
- 13:25
- Unidentified speaker (UNKNOWN): If
- 13:25
- Unidentified speaker (speaker_0): those are the strategies, it's actually a subgame perfect Nash equilibrium, incidentally. Right? That it's important to say, to just a to think through and spell out that given, let's say with a 253, if all the lines lions as part of their strategy know that yeah, if we ever find ourselves in a scenario where there's 252 of us looking at either
- 13:45
- Unidentified speaker (speaker_0): a steak or a sleeping lion, everyone decides to just watch them. And by the way, the and the reason that's the rational response is that you have to technically, and this is what I meant when I said it's what's called a sub game perfect equilibrium where everybody knows like they check and just see what if we did something different
- 14:05
- Unidentified speaker (speaker_0): and that it just it keeps being that everyone's doing something rational given what everyone else's strategy is all the way down. So if everybody's actual strategy is to say, if I'm ever part of a group of lines that's an even number, I'm not gonna eat the thing in front of me. But if an odd number, I'm gonna going to try to eat the thing if I'm there first.
- 14:25
- Unidentified speaker (UNKNOWN): If
- 14:25
- Unidentified speaker (speaker_0): that's what all the lines are doing, then that's all mutually consistent and compatible and supports the Nash equilibrium, right? Because again, if you're the 253rd lion, you go ahead and eat the steak because you know I'm safe. So you benefit from doing that, whereas
- 14:45
- Unidentified speaker (speaker_0): there's no downside because every you know what everyone else is doing. All right. Again, also the reason when there's 252 lions looking at the lion taking a nap, the reason it is a best response for each of them to say, I'm not going to eat that line that's taking a nap, even though, oh, that would give me short-term gain, I
- 15:05
- Unidentified speaker (speaker_0): get a meal, is because they know everybody else's strategy chosen
- 15:10
- Unidentified speaker (UNKNOWN): in
- 15:11
- Unidentified speaker (speaker_0): this equilibrium is to say, if for some reason the first guy, the first lion ate the steak and is taking a nap, and then for some reason,
- 15:22
- Unidentified speaker (UNKNOWN): when
- 15:22
- Unidentified speaker (speaker_0): there's 252 of us, one of them ate that thing at that point, it would be safe for me to eat him, and that's what I would do.
- 15:31
- Unidentified speaker (UNKNOWN): All right,
- 15:31
- Unidentified speaker (speaker_0): because I would look ahead and know that if we get down to 250, then we're fine.
- 15:37
- Unidentified speaker (UNKNOWN): All
- 15:37
- Unidentified speaker (speaker_0): right. So that's why it's all supported. The first line knows, or the 253rd, let's say, to avoid confusion, knows I can go ahead and eat that steak because everybody else, they won't. And the reason it's rational for them not to, when they're looking at me taking a nap, is because they look ahead one step themselves and say that's what I know what everyone
- 15:57
- Unidentified speaker (speaker_0): else is going to do if I were to do that. And so it all supports it. It's all consistent. Everybody does what is expected. Everybody Everything makes sense. It all works out. And everybody is best responding given what everybody else's strategy choice is. Right? That's what it means to be a Nash equilibrium. Okay. So that's what they wanted the official answer to be, I'm quite sure.
- 16:16
- Unidentified speaker (UNKNOWN): And
- 16:16
- Unidentified speaker (speaker_0): you can see why some people are saying, What the heck does this have to do with financial portfolio management or whatever? And okay, but I think it's pretty obvious why Goldman Sachs would be. curious just to see what a candidate say on a question like this. Incidentally, somebody that I know who is on Twitter, he saw this and he said, I used to work for
- 16:36
- Unidentified speaker (speaker_0): Goldman when he would I don't know if he was actually doing the interviews or he was just in HR and knew the kind of questions they had. So he said, I can't confirm that I ever saw this literal question, but he said, Yeah, this is definitely we would ask stuff like this. you know, or at least back when I was overlapping with the job interviews.
- 16:56
- Unidentified speaker (speaker_0): Okay, so I think we can assume that this is if this is not literally a Goldman Sachs question, that certainly this is not stretching to say that this is the kind of thing they ask. Okay, so my problem with this, I'll spell it out here, but let me just say this is something that I've had
- 17:14
- Unidentified speaker (UNKNOWN): a
- 17:14
- Unidentified speaker (speaker_0): problem. with for a while. So I'm gonna link at the show notes page on a different podcast that I often co-host with this guy, Adam Heyman, I don't know, about a year ago or something. We went through
- 17:26
- Unidentified speaker (UNKNOWN): the
- 17:26
- Unidentified speaker (speaker_0): issues with what's called the finitely repeated prisoner's dilemma. And so the point I made there is the same type of point that I'm gonna make in response to this alleged correct solution to the Goldman Sachs question.
- 17:39
- Unidentified speaker (speaker_0): Okay. And so the the prisoner's dilemma, I'm sure most of you have heard of a prisoner's dilemma where the payoffs are such that if both people cooperate, they each get a high payout, but the very highest payout occurs where you defect and the other guy cooperates, but then knowing that
- 17:58
- Unidentified speaker (UNKNOWN): if
- 17:58
- Unidentified speaker (speaker_0): they both defect, they get a worse outcome than if they both cooperated. All right. And so it's strategically, it's a situation where people acting in their narrow self-interest end up getting a worse outcome than if they did what prima facie looks like it's against their interest,
- 18:17
- Unidentified speaker (UNKNOWN): but
- 18:17
- Unidentified speaker (speaker_0): it's the kind of thing where, oh, but if other people do it too, then we're all better off. Right. And so I think that's a good model of ethical behavior. Like if we all live in a society where people refrain from theft, we're all better off than if we all steal whenever we can get away with
- 18:36
- Unidentified speaker (UNKNOWN): it.
- 18:37
- Unidentified speaker (speaker_0): Right. But given whatever everyone else is doing, your narrow interest, your pecuniary interest, you're always better off if you steal when you can get away with it.
- 18:47
- Unidentified speaker (UNKNOWN): Now
- 18:48
- Unidentified speaker (speaker_0): you could say, well, what if somebody sees me and I go to jail? or would they retaliate? Sure. But again, I'm saying if you could get away with it.
- 18:56
- Unidentified speaker (UNKNOWN): And
- 18:56
- Unidentified speaker (speaker_0): you could say, oh, that's why we got to set up monitoring and stuff. And that's true. But in general, there's plenty of times in life, like if you were over at somebody's house party or something, you could probably steal something and they wouldn't be sure it was you. And as long as you didn't do it too much, and you could do that and you'd probably be able to gain some money over. the course of a year in
- 19:16
- Unidentified speaker (speaker_0): a pretty safe way. And I'm saying we are better off, all of us, given that most of us don't do that. Just be even though you could like, why would you do you know, I'm not gonna do that? All right. So I think that's the prisoner's dilemma actually captures situations like that. I mean, in its literal context, if you had two people that were brought in and the
- 19:36
- Unidentified speaker (speaker_0): cops put them in separate interrogation rooms and say, Oh yeah. the other guy already told us everything and you're going away for 30 years. However, if you admit what you did and just write out this confession, we'll knock it down to 10.
- 19:48
- Unidentified speaker (speaker_0): Especially like if the cops are lying, right? So the idea is if both of those people just know, nope, I'm not ratting out my buddy, they both do better than if they take the plea, even though you know the prosecutors look at we'll get you out of here with probation if you just admit what. you did and pin it on your partner here. But then if they both do that, then they both end up with worse than if they both just kept
- 20:08
- Unidentified speaker (speaker_0): their mouth shut. Okay. So you can see why that, I guess, it's a weird thing because is it really ethical is two bank robbers get brought in and like, no no, no, I'm not gonna testify against my partner if they're bank robbers, then it's not obvious that that's the noble thing to do, but you get the idea. Okay. So anyway,
- 20:25
- Unidentified speaker (UNKNOWN): that's
- 20:25
- Unidentified speaker (speaker_0): the prisoner's dilemma.
- 20:27
- Unidentified speaker (speaker_0): where it seems that doing what's in your narrow self-interest actually leads to
- 20:32
- Unidentified speaker (UNKNOWN): a
- 20:32
- Unidentified speaker (speaker_0): worse outcome for you than if you did the quote sucker move by cooperating, so long as you thought the other guy might do it too. So what's interesting is even though you can't make that work in a one-shot example, if you have if you repeat it an infinite number of times, that same game and you're playing the same percentage. person, you can
- 20:52
- Unidentified speaker (speaker_0): support cooperation.
- 20:55
- Unidentified speaker (UNKNOWN): Like
- 20:55
- Unidentified speaker (speaker_0): if your rule is, oh, I'm going to start cooperating,
- 20:59
- Unidentified speaker (UNKNOWN): and
- 20:59
- Unidentified speaker (speaker_0): then if this guy does too,
- 21:01
- Unidentified speaker (UNKNOWN): then
- 21:01
- Unidentified speaker (speaker_0): I'll just keep cooperating, right? But if he doesn't, if he defects,
- 21:07
- Unidentified speaker (UNKNOWN): even
- 21:07
- Unidentified speaker (speaker_0): if it's like in round 13,
- 21:09
- Unidentified speaker (UNKNOWN): well,
- 21:09
- Unidentified speaker (speaker_0): then from that point on, I'm going to defect. You can have rules like that, or another popular one is what's called tit for. tat, where you just you start out cooperating and then you just mimic what for each round you mimic what the other guy did the prior round. So if you both start cooperating and you're both doing that overall meta strategy, you just cooperate forever. And so the idea is
- 21:29
- Unidentified speaker (speaker_0): you rack up a ton of points
- 21:32
- Unidentified speaker (UNKNOWN): when
- 21:32
- Unidentified speaker (speaker_0): you're both cooperating.
- 21:34
- Unidentified speaker (UNKNOWN): All right.
- 21:35
- Unidentified speaker (speaker_0): And so if you could just get long strings of cooperation, you both earn more than if you just each do the narrow, quote, optimal, rational thing of defecting all the time.
- 21:45
- Unidentified speaker (UNKNOWN): All
- 21:45
- Unidentified speaker (speaker_0): right. And then they've actually done like computer simulations and tournaments in these types of things where different agents go in with different types of strategies and they interact with each other. And then the afterwards they just see who racked up the highest score. And the agents that go in thinking, no, I'm going to do the game thing theoretic, rational thing, and just defect
- 22:05
- Unidentified speaker (speaker_0): all the time, they don't win the tournament. So anyway, I'll link to a sh an episode where I really walk through that stuff and show you the math. But my takeaway there was even though in the long run, if it's an infinitely repeated one, you can support cooperation mathematically with all the bells and whistles. If it's just finitely repeated, you can't,
- 22:25
- Unidentified speaker (UNKNOWN): because
- 22:25
- Unidentified speaker (speaker_0): you can always go ahead. to the last round and just say clearly when we get to the last round, even if it's two billion rounds of the prisoner's dilemma with the same guy over and over, I go to the two billionth round at that point in bygones or bygones, doesn't matter what happened before. Now this guy and I are playing one last prisoner's dilemma, and we know the quote rational
- 22:45
- Unidentified speaker (speaker_0): thing to do is to defect.
- 22:47
- Unidentified speaker (UNKNOWN): So
- 22:47
- Unidentified speaker (speaker_0): we know in that last round we're both going to defect,
- 22:50
- Unidentified speaker (UNKNOWN): but then
- 22:51
- Unidentified speaker (speaker_0): that means the second last round, there's no advantage to cooperating because the I know my opponent or the other guy, don't call him opponent because that is biased, my co-player,
- 23:03
- Unidentified speaker (UNKNOWN): is
- 23:03
- Unidentified speaker (speaker_0): going to defect in the last round. And so clearly I should then defect in the second last round. But he thinks that too, and you just use that backwards induction logic to know, oh, we're going to gonna both defect all the way up to the first round. And so here, I'm now linking it to the line thing. And I, if you want to see it spelled out in more
- 23:23
- Unidentified speaker (speaker_0): detail, I'll link to that episode in the show notes page. But my point is, there
- 23:29
- Unidentified speaker (speaker_0): it's it seems like airtight logic to say, oh, yeah, so that's why both players are gonna going to just defect forever. But I'm saying, what if you started playing the game with? somebody and you're defecting and defecting and they just keep cooperating?
- 23:42
- Unidentified speaker (UNKNOWN): And
- 23:43
- Unidentified speaker (speaker_0): you're like, what the heck? And so maybe you're just going to keep doing it because you think, yeah, this guy's just nuts and I don't know what he's doing. But
- 23:50
- Unidentified speaker (UNKNOWN): if
- 23:50
- Unidentified speaker (speaker_0): you thought maybe he does have some kind of rule that says, I'm going to try cooperating for the first 20 times
- 23:56
- Unidentified speaker (UNKNOWN): to
- 23:56
- Unidentified speaker (speaker_0): see if I can get this guy
- 23:58
- Unidentified speaker (UNKNOWN): to
- 23:59
- Unidentified speaker (speaker_0): go along with me, because if we can just get a long stretch of cooperation mutually, we're both going to be so much better off than if we just sit there and do the safe thing, the defensive thing of just defecting the whole time.
- 24:11
- Unidentified speaker (UNKNOWN): And
- 24:11
- Unidentified speaker (speaker_0): so I'm gonna give this guy a chance to learn and see what I'm doing. Maybe he'll get the hint. And so I'm saying I think it would be irrational if you saw some guy who just keeps cooperating and you say, Nope, nope. I know that the unique, what's called rationalizable outcome. for finite repeated prisoners dilemma, the certainly the unique Nash
- 24:30
- Unidentified speaker (speaker_0): equilibrium,
- 24:31
- Unidentified speaker (UNKNOWN): and
- 24:31
- Unidentified speaker (speaker_0): there's all sorts of other, you can say much stronger things, is that we just defect the whole way through. So I'm just going to keep defecting because I would be a sucker if I cooperated because I know he's going to just defect every single round. And how could you keep thinking that
- 24:45
- Unidentified speaker (UNKNOWN): as Is
- 24:45
- Unidentified speaker (speaker_0): you keep seeing him cooperate round after round? So that's the argument I spelled out. If you want to see it, I just kind of gave you the punchline.
- 24:51
- Unidentified speaker (speaker_0): I'll link to that. But so now back to this line one, it's a similar issue. Is let's say that there's 252 lions. All right. So I'm changing the question. So what the Goldman people, the reason they picked an odd number that was really high is they want you to say, Oh yeah, that first, that lion's gonna eat it, and then the rest are gonna
- 25:11
- Unidentified speaker (speaker_0): sit there and look at him. them. But suppose the question were 252. well then the quote correct answer
- 25:17
- Unidentified speaker (UNKNOWN): would
- 25:17
- Unidentified speaker (speaker_0): be to say, oh, all the lions would just sit there and look at it and not do anything. Because they all know, being good game theorists, that if you're ever in a situation where there's 251 lions looking at a meal,
- 25:32
- Unidentified speaker (UNKNOWN): the
- 25:32
- Unidentified speaker (speaker_0): one lion's Is gonna going to eat it,
- 25:34
- Unidentified speaker (UNKNOWN): right?
- 25:34
- Unidentified speaker (speaker_0): And so that's why if there's 252, nobody's touching the thing.
- 25:38
- Unidentified speaker (UNKNOWN): And
- 25:38
- Unidentified speaker (speaker_0): I say, well, hang on. Let's just think that through. Let's poke at that. So suppose we are in that situation and I'm the 252nd lion. And I said, I'm going to go ahead and eat that steak.
- 25:49
- Unidentified speaker (UNKNOWN): And we
- 25:50
- Unidentified speaker (speaker_0): said, What are you doing? Why would you do that? Because you know
- 25:54
- Unidentified speaker (UNKNOWN): the
- 25:54
- Unidentified speaker (speaker_0): 251 now looking to you taking a nap are going to eat you, the first one that gets to you, because they know if we get to 250,
- 26:01
- Unidentified speaker (UNKNOWN): they're
- 26:01
- Unidentified speaker (speaker_0): not going to gonna touch. him. Because we all know if there's an even number of lines, nobody touches anything.
- 26:08
- Unidentified speaker (UNKNOWN): And
- 26:08
- Unidentified speaker (speaker_0): I say, wait a minute though, wouldn't I have introduced doubt into that belief
- 26:13
- Unidentified speaker (UNKNOWN): if
- 26:14
- Unidentified speaker (speaker_0): when there's 252, I eat the steak?
- 26:17
- Unidentified speaker (UNKNOWN): Does
- 26:17
- Unidentified speaker (speaker_0): everyone see what I'm saying? How is it rational? How is it perfectly rational
- 26:22
- Unidentified speaker (UNKNOWN): if
- 26:22
- Unidentified speaker (speaker_0): the 252nd line eats the steak, he's out, and now there's 251.
- 26:27
- Unidentified speaker (speaker_0): And they all think, oh, clearly, whoever can get to that line first should eat him because we're totally safe then. Yeah, we're going to fall asleep, but we are dead set certain that if there's 250 looking at me, nobody would be so foolish as to eat me. How do you know that? Because you just saw that the 252nd
- 26:47
- Unidentified speaker (speaker_0): line did eat somebody or the steak. And so notice here, I some people, when I said that on Twitter, and I was trying to argue that no, I disagree that this alleged solution is as airtight as people think it is. Some people were pushing back and they were just saying, Well, okay, but you're just then deviating from the question that it says all lines are perfectly rational. And you're just saying,
- 27:08
- Unidentified speaker (speaker_0): well, no, I'm going to suppose they're not. And yeah, you're going to gonna get a different answer if you don't do what the question said and the premises. No. I understand. that reaction, but I'm pushing back and saying, no, that's not what I'm doing. I'm saying the alleged official answer doesn't have the lines being perfectly rational. They are in that one particular equilibrium, like that's internally
- 27:27
- Unidentified speaker (speaker_0): consistent. Those beliefs all fit together and mesh. But I'm saying that's not the only possible outcome that would be consistent with all the lines being perfectly rational. That's what I'm saying.
- 27:40
- Unidentified speaker (UNKNOWN): And
- 27:40
- Unidentified speaker (speaker_0): so, in particular,
- 27:42
- Unidentified speaker (UNKNOWN): like
- 27:42
- Unidentified speaker (speaker_0): I said, if I'm walking through a scenario and say, okay, why might, you know, again, start with 252 because it's easier for me to make my case there.
- 27:49
- Unidentified speaker (UNKNOWN): If
- 27:50
- Unidentified speaker (speaker_0): there's 252 lions, why are we so sure that they it would be suicidal for any of them to eat it? That they might do it thinking, oh, then because I've introduced doubt, now when there's 251 looking at me because I'm taking a nap, they might think, uh, there's a chance if I eat him
- 28:10
- Unidentified speaker (speaker_0): and I'm asleep, then of the 250, someone might eat me because they will say, no, I just saw the 252nd line do it. So my point is it's not rational. And you could keep pushing it, right? Like what if the line lion just keeps keep doing it? And what if we get down to
- 28:25
- Unidentified speaker (UNKNOWN): there's
- 28:26
- Unidentified speaker (speaker_0): 51 fifty one line? lions and they just saw two hundred lions get eaten, two hundred and what in one or two? There's fifty one left now, and there's a lion sleeping there. And so are we still so sure that to say, Well, yeah, that
- 28:43
- Unidentified speaker (UNKNOWN): they
- 28:43
- Unidentified speaker (speaker_0): should go ahead and eat that lion because everybody knows of the 50 fifty remaining, nobody would touch me now because every everybody knows there's an even number of lines, they don't eat anything. Even though you just saw with your own eyes that rule get violated a hundred times in a row, right? That doesn't make any sense. That's not rational. How is that rational?
- 29:03
- Unidentified speaker (UNKNOWN): Okay.
- 29:03
- Unidentified speaker (speaker_0): And so that also, just to go back, I think sometimes people push back and they say, well, sure. We realize in the real world, people fall short of perfect. rationality. We're just holding that up as an ideal, knowing that, yeah, people aren't going to live up to just like you can like define what the perfect chess
- 29:23
- Unidentified speaker (speaker_0): play is, even though even two grandmasters going at it are going to screw up relative to what the chess engine would recommend that they do. Okay. But no, that's not what's going on here. I just want to stress that, right? That start, you know, back to the original thing is suppose there's 252. I'm not merely saying, oh, what if the 252nd line made a
- 29:43
- Unidentified speaker (speaker_0): mistake and ate the stake? Then the next one also makes a mistake or is worried about other people making mistakes and whatever, and da-da-da. And so yeah, you could get some weird thing that differs from what the official answer is because some lines make mistakes. It's not what I'm saying. Because if
- 30:00
- Unidentified speaker (UNKNOWN): when
- 30:01
- Unidentified speaker (speaker_0): there's 252. if that lion, you know, the first one of the steak eats it, thinking it's possible that when there's 251, even though using the backwards induction logic, you would think I'm a dead, I'm a goner for sure. Maybe not. Because again, the fact that we've
- 30:21
- Unidentified speaker (speaker_0): now broken the spell and shown you can't conclude that if there's an even number, nobody's eaten it. because I just did it. I was a 250-second line and I ate that steak. And so maybe now because I've shattered that illusion,
- 30:34
- Unidentified speaker (UNKNOWN): well,
- 30:34
- Unidentified speaker (speaker_0): I'm taking a nap and there's 251 lines looking at me, maybe they'll look ahead one round and say, Well, if I eat him, I'm out, and there's 250.
- 30:42
- Unidentified speaker (UNKNOWN): And
- 30:42
- Unidentified speaker (speaker_0): I just saw with my own eyes that it is possible, even if there's an even number, that eating occurs. And so maybe I won't eat them. And so then that guy takes an ab and wakes up. So notice how can you say he made a mistake
- 30:56
- Unidentified speaker (UNKNOWN): being
- 30:57
- Unidentified speaker (speaker_0): the 252nd line when he ate the steak and survived? Right. And so just to flip back to the finally repeated prisoner's dilemma, when I would make these points years ago, I would get pushback from some conventional game theorists and they would say, Well, yeah, well, we're not predicting that in the real world. actual human beings playing a repeated
- 31:17
- Unidentified speaker (speaker_0): prisoner's dilemma, but that's finite. And by the way, in case you're saying, well, of course they couldn't play an infinite, but what you could do is like if they flip a coin after each round, and if it's heads, they play another one, if it's tails, they stop. That has the same flavor
- 31:30
- Unidentified speaker (UNKNOWN): as
- 31:31
- Unidentified speaker (speaker_0): a finite repeated prisoner dilemma where you have a discount rate on the future so that the payoffs you get from round 10,000 ten thousand don't weigh as heavily today as the payoff you get from today's round. So that also you can have the same strategic flavor of an infinitely repeated game if two players are sitting there and they just
- 31:51
- Unidentified speaker (speaker_0): they flip a coin to decide do are we gonna keep going? So the point is the critical thing is that
- 31:57
- Unidentified speaker (UNKNOWN): in
- 31:57
- Unidentified speaker (speaker_0): the beginning they can't look ahead and know when the final round is gonna going to occur.
- 32:02
- Unidentified speaker (UNKNOWN): So like
- 32:02
- Unidentified speaker (speaker_0): if there's a random event that happens to determine are we going to play another round or not, that means at the outset,
- 32:07
- Unidentified speaker (UNKNOWN): you
- 32:08
- Unidentified speaker (speaker_0): don't know how many rounds there will be. So you can't go to the last round and say, well, clearly if we get to round 2007,
- 32:14
- Unidentified speaker (UNKNOWN): we
- 32:14
- Unidentified speaker (speaker_0): know we're both going to defect. So that means around 2006, right? So to do that backwards logic, you need to know how many rounds are there going to be. All right. Incidentally, in case you're getting tripped up on that. So my point being there, when I would make these points that I've been doing. here, some of the conventional game theorists would push back and say, Yeah, yeah, we know in a fight on the repeated prisoner's dilemma, real humans aren't going to defect every
- 32:34
- Unidentified speaker (speaker_0): round. We're just saying they ought to. If they were perfectly rational, they would.
- 32:38
- Unidentified speaker (UNKNOWN): And
- 32:38
- Unidentified speaker (speaker_0): so I want to say, okay, so you got two different teams. Team A has two game theorists playing. And we're gonna what we're gonna do is play the prisoner's dilemma a thousand rounds in a row. Both teams know that.
- 32:51
- Unidentified speaker (UNKNOWN): The
- 32:51
- Unidentified speaker (speaker_0): two game theorists defect the whole time
- 32:54
- Unidentified speaker (UNKNOWN): and
- 32:54
- Unidentified speaker (speaker_0): they get their payoff. The average people from the street are playing and they cooperate for 850 rounds. Then on the 851st round, one of the guys defects, the other guy cooperated.
- 33:07
- Unidentified speaker (UNKNOWN): That
- 33:07
- Unidentified speaker (speaker_0): keeps happening for a little bit, and then they both just defect from round 903 to 1,000. And lo and behold, the two Joes from the street are going to have way higher payoffs overall than the two game theorists.
- 33:21
- Unidentified speaker (UNKNOWN): So
- 33:21
- Unidentified speaker (speaker_0): I'm going to say, in what sense could the game theorists say, Oh, you idiots, you're clearly not as rational as we are, when they got a higher score on the game than you did. It doesn't make any sense. Right. So likewise here, I'm saying it doesn't work to say, oh, yeah, if the Lions aren't rational, anything goes when they get a higher payoff from being quote
- 33:41
- Unidentified speaker (speaker_0): irrational. Okay, so
- 33:43
- Unidentified speaker (UNKNOWN): I
- 33:44
- Unidentified speaker (speaker_0): think I've spelled out my point. Let me in closing is clarify something. And again, it kind of goes back to what I said at the beginning, or near the beginning, about the different layers and that you have to know that they know that they know, and you gotta keep doing
- 34:00
- Unidentified speaker (UNKNOWN): that.
- 34:00
- Unidentified speaker (speaker_0): So
- 34:01
- Unidentified speaker (UNKNOWN): back
- 34:01
- Unidentified speaker (speaker_0): to if there's one lion and we say, oh, he's rational, all we mean is the lion understands that, oh, eating a steak gives me benefits and there's no downside, so I'm gonna do it. Right. So there, rationality just means the lion is gonna be intelligent enough to eat the steak when there's no
- 34:21
- Unidentified speaker (speaker_0): downside. Okay.
- 34:23
- Unidentified speaker (UNKNOWN): If
- 34:23
- Unidentified speaker (speaker_0): there's two lions to say they're perfectly rational in the game theoretic sense, and you know what the standard game theorist means by that, and this kind of a problem, I'm also okay with. Because all it means is each lion can look ahead one round and say, All right, if I eat that steak, I'm
- 34:43
- Unidentified speaker (speaker_0): out. And then it's just one lion looking at me. I predict he would eat me in that situation. And so they there, all you're attributing to that line is the capacity for thought of thinking, would I eat a passed out line if there's no downside? Yes, I would. All right. So that I'm fine with
- 35:03
- Unidentified speaker (speaker_0): that. But when it goes to the next one, now it gets dicier because it's now what you have to do, right? So if there's so again, I'm just to be clear, I'm saying if there's two lions now. I'm perfectly fine saying perfect rationality entails that they're not gonna neither of them will touch the stake. But now when
- 35:23
- Unidentified speaker (speaker_0): there's three,
- 35:24
- Unidentified speaker (UNKNOWN): it
- 35:24
- Unidentified speaker (speaker_0): gets more dubious because here
- 35:27
- Unidentified speaker (UNKNOWN): you
- 35:27
- Unidentified speaker (speaker_0): gotta say, oh, if I eat that, then there's gonna be two. And so it's not enough to know if there's three of them looking at the thing and saying, Okay, it's safe to eat. You have to know what the lines are gonna be doing, how they're gonna reason around if there's two of
- 35:47
- Unidentified speaker (UNKNOWN): them.
- 35:47
- Unidentified speaker (speaker_0): Right. And you have to know that that they know that as well.
- 35:51
- Unidentified speaker (UNKNOWN): Okay.
- 35:52
- Unidentified speaker (speaker_0): And so you can see how each time you build it up, it's getting more and more complex. And again, that chain is then fragile because you realize, or you could realize.
- 36:03
- Unidentified speaker (speaker_0): well, what if I deviate from that? And I demonstrate that this chain that we've built up through induction clearly isn't holding. And then by doing so, what if that means then the move you made that ostensibly was foolish ends up actually being wise? Well, then why wouldn't you do that? And so
- 36:23
- Unidentified speaker (speaker_0): again, to like it would reinforce for everybody, like, oh, the reason that probably person did that, even though
- 36:29
- Unidentified speaker (UNKNOWN): a
- 36:29
- Unidentified speaker (speaker_0): minute ago we thought that was a silly thing to do,
- 36:33
- Unidentified speaker (UNKNOWN): the
- 36:33
- Unidentified speaker (speaker_0): reason that person did that is because actually, now that we're recalibrating our expectations, he might get away with it. And so again, it wasn't even just a mistake on his part, right? It's a bold move. Let's see if it pays off.
- 36:47
- Unidentified speaker (UNKNOWN): Okay,
- 36:48
- Unidentified speaker (speaker_0): so anyway, that's I just want to clarify there that the further away you get from ground zero, when you talk about all the lines being perfectly rational, it gets more and more dubious because it's not just that they know all the other lines are rational. It's that they have to all think that, oh, we're all reasoning about this scenario strategically
- 37:07
- Unidentified speaker (speaker_0): the same way, and we're all in agreement with each step of the argument. You need all that to be packed in as well.
- 37:15
- Unidentified speaker (speaker_0): Okay. Let me um actually end. I forgot I wanted to mention if you've never heard of it, this is a fun one. There's this thing called the paradox of the unexpected hanging. So I'll link to it
- 37:26
- Unidentified speaker (UNKNOWN): in
- 37:26
- Unidentified speaker (speaker_0): the show notes page if you want to go see like an official treatment of it somewhere. But this was something that was making the rounds
- 37:33
- Unidentified speaker (UNKNOWN): in And
- 37:33
- Unidentified speaker (speaker_0): like a bunch of mathematicians and logicians and things like that, philosophers were already arguing about this.
- 37:40
- Unidentified speaker (UNKNOWN): I
- 37:40
- Unidentified speaker (speaker_0): want to say like in the 1940s or 50s. I think this is when this was big. But anyway, here's the quick story. I hope I can it's been a while since I've read about it. I'll try to do it justice. But you'll see the flavor of this is similar to what we just went through.
- 37:55
- Unidentified speaker (UNKNOWN): So
- 37:55
- Unidentified speaker (speaker_0): the story goes: there's some guy who gets convicted of murder, and the judge is a bit quirky. And so the guy comes to hear his sentence and the judge says, Okay, it is Sunday morning right now. And you're gonna go back to your cell, and then every morning at precisely 9 a.m., right now it's a little bit before
- 38:15
- Unidentified speaker (speaker_0): nine, the guards are gonna come and tell you whether you're gonna be hanged that day or not. Okay. And what I am telling you is you will be hanged.
- 38:27
- Unidentified speaker (speaker_0): at some point during this week, and when it happens, you will be surprised. All right.
- 38:34
- Unidentified speaker (UNKNOWN): And
- 38:34
- Unidentified speaker (speaker_0): so that's the thing. And so the guy goes back to his cell and he's just like, oh, geez, and he's really miserable because he knows he has at most a week to live. He's gonna get his affairs in order and hope to talk to his loved ones and say goodbye before that event befalls him. But then his lawyer shows up and he says, Oh, this is great. And the guys, what are you talking about? This is
- 38:54
- Unidentified speaker (speaker_0): awful. He goes, No, no. I was thinking it through and listen.
- 38:57
- Unidentified speaker (UNKNOWN): He
- 38:57
- Unidentified speaker (speaker_0): says, Today's Sunday, let's say we get through, Friday night comes and goes, and that means you wake up Saturday and seven o'clock in the morning, and you know they're going to come soon to give you the verdict. He says, At that point, if you survived up until then and the judge told you you're going to. be hanged this week, so you know
- 39:17
- Unidentified speaker (speaker_0): it's got to be Saturday, right? So then when they come and tell you that yep, today's the day we're doing it, you won't be surprised. Because you'll know, like as a Friday night, or technically even Friday, you know,
- 39:30
- Unidentified speaker (UNKNOWN): 10
- 39:31
- Unidentified speaker (speaker_0): a.m. or whatever, you know, the minute after I said whatever the ver you
- 39:33
- Unidentified speaker (UNKNOWN): know,
- 39:33
- Unidentified speaker (speaker_0): they would tell you if I'm getting mixed up, is what I said.
- 39:36
- Unidentified speaker (UNKNOWN): You
- 39:36
- Unidentified speaker (speaker_0): would know at that point, well, it's gotta got be to Saturday then. And so then when they actually go to tell you that on Saturday, you're not gonna be surprised.
- 39:42
- Unidentified speaker (UNKNOWN): He
- 39:42
- Unidentified speaker (speaker_0): said, But the judge told you when we inform you, you're gonna be surprised. So I said, Yeah. And he said, So clearly it can't be Friday night. This isn't gonna last. You're not gonna be still here Friday night because that means they would have to tell you on Saturday, but then you wouldn't be surprised, right? So we can just rule that out. And guys I said, like, yeah, okay.
- 40:01
- Unidentified speaker (UNKNOWN): And the
- 40:02
- Unidentified speaker (speaker_0): guy says, Oh, okay. But now, suppose it gets to be Thursday night and you're still alive. Yeah.
- 40:09
- Unidentified speaker (UNKNOWN): He
- 40:09
- Unidentified speaker (speaker_0): says, So what does that mean? He says, We know they can't hang you on Saturday. So that would mean they have to hang you on Friday. But then that means Friday morning when the guard comes to tell you the news and he says, Yeah, it's today. You're not going to be surprised. You already knew that.
- 40:25
- Unidentified speaker (UNKNOWN): And I
- 40:25
- Unidentified speaker (speaker_0): was like, Yeah, that's right.
- 40:26
- Unidentified speaker (UNKNOWN): He
- 40:26
- Unidentified speaker (speaker_0): says, I can't hang you on Thursday. Or sorry, Friday.
- 40:30
- Unidentified speaker (UNKNOWN): He
- 40:30
- Unidentified speaker (speaker_0): goes, Yep, that's right. Okay. And then he just keeps walking him through that logic until it gets to the point, you know, it can't be Thursday, it can't be Wednesday, it can't be Tuesday, it can't be Monday. So now it's Sunday. And like the guard's gonna come in a few minutes to tell him. And he's gonna say, So it's gotta be today, but it can't be because you know that.
- 40:49
- Unidentified speaker (UNKNOWN): Right.
- 40:49
- Unidentified speaker (speaker_0): And so if he tells you, you're not gonna be surprised because you know it's gotta be today, right? And so the judge would be, you know, his prediction would be wrong.
- 40:55
- Unidentified speaker (UNKNOWN): And he goes,
- 40:56
- Unidentified speaker (speaker_0): Yeah. So he says, So clearly, we've just walked through
- 40:59
- Unidentified speaker (UNKNOWN): there
- 40:59
- Unidentified speaker (speaker_0): can't be any day at which the guard tells you today's the day that you're gonna get hanged, because otherwise the judge's prediction doesn't come true.
- 41:08
- Unidentified speaker (UNKNOWN): And he
- 41:08
- Unidentified speaker (speaker_0): says, Yeah, you're right. He says, See, great.
- 41:11
- Unidentified speaker (UNKNOWN): So the
- 41:11
- Unidentified speaker (speaker_0): guy's just sitting there, doot doo doot, he's having a blast because he knows I'm totally safe. And then Sunday gets true, the guard says it's not today. And he said, ha ha ha. And the lawyer, you know, gives him a wink. And then Monday comes, the guard comes and says, Not today. The guy said, Uh, I didn't think it would be. Tuesday, same deal. Wednesday, the guard
- 41:31
- Unidentified speaker (speaker_0): says, Yep, today's the day we're hanging you. And the guy's shocked. And so he gets hanged, and everything the judge said is perfectly true.
- 41:39
- Unidentified speaker (speaker_0): Discuss. All right. See you next time, folks.
- 41:43
- Unidentified speaker (speaker_1): This concludes another episode of InFi, The Future of Finance, with Dr. Robert Murphy. The information provided is for educational purposes and does not constitute financial advice. Consult with qualified professionals before making any financial or investment decisions. For more information on the host and for previous episodes, visit
- 42:03
- Unidentified speaker (speaker_1): infinio.ai.
- 42:06
- Unidentified speaker (speaker_1): Thanks for listening.