ControlsFreak
No bio...
User ID: 1422
ControlsFreak
1d ago
What employees have you fired so far?
In Zvi's recent post, I noticed an interesting pairing of two things:
Sell your house. Stuart Thompson lets Gemini (because he had a free account there from work that saved him $8 a month?!) walk him through everything involved in the sale, including being his agent. The problem is, Stuart does not seem to realize he does not know the counterfactual?
Stuart A. Thompson: In the end, using A.I. netted me more than $90,000. That includes the premium over the asking price, plus the roughly $36,000 in fees I didn’t pay.
I mean, yes, the agents he talked to early on told him he’d lose money, and instead he turned a profit. But only after the sale did he talk to another agent for an expert opinion, and that expert expected a higher sale price than Stuart got, meaning he almost certainly listed too low. Stuart thinks that after the agent fee he still basically broke even, but I’m guessing he put in more work and stress this way, and took on more downside risk. I know that if I am ever selling or buying, I will be using AI extensively as part of the effort, but I am going to stick with Danielle Wiedemann. I am confident that her help, connections and advice were worth far more than the fee, and would be again.
and:
For those confused about the radiology example, yes, AI is better than radiologists at reading x-rays, and many other components of professional services, and does so at cost epsilon, and this is super useful. Even if no one is out of work quite yet, often there is a ton of value in ‘pretty good answer, vastly better than you could otherwise get without a professional, for cost ~$0’ when the professional costs $1,000 and up.
It was a bit stark, because getting a pretty good answer, vastly better than you could otherwise get without a real estate professional, would seem to cost ~$0, when the professional apparently costs something like $36,000 and up. So why not fire the real estate agent?
There could be a variety of reasons involving the nature of the work, regulatory barriers, etc., but one thing that comes to mind is that Zvi has paid for a real estate agent before and is consciously thinking about what that situation is like when thinking about whether he would hire again. Whereas, I doubt he's hired a radiologist before and is probably not in a situation where he's thinking super seriously about the considerations that would be involved if he had a need for such a service.
This leads me to ask, "Which employees have you fired?" In this case, "employees" can be read broadly, covering folks like real estate agents/radiologists, who you may procure services from on occasion, in addition to actual employment relations if you're a manager/business owner. But I want to particularly hone in on examples where you have paid a human for a particular service in the past and have subsequently encountered a nearly-identical need, but have chosen to not pay a human now for the service.
This question is in significant part simply selfish. I might be missing some aspect in my life where I can save a bunch of money. That would be cool, and I'd like to do that if I can.
ControlsFreak
1d ago
The burns example where your buddy is in horrible pain and bound to die soon is another one that works. You can play with it by having him be actively begging for death or just screaming wordlessly.
It doesn't meet the criteria stated above:
I see no circumstances under which the principle "Don't murder innocents" must be compromised in order to live. [emphasis added]
That may not be the only category in which some folks think something is acceptable, but it is the criteria stated that I'm comparing to for purposes of this sub-thread.
ControlsFreak
1d ago
Almost none of your examples actually work. Most of them get intentionality the wrong way 'round. There is obviously a huge conceptual chasm between an affirmative requirement to take extreme measures to save a life and a far more minimal requirement that one not murder. Perhaps you're just confused about what 'murder' is? Or maybe about what "in order to live" means?
Your buddy falls while climbing and you have to cut the rope so you both don't die.
This is the only one that actually gets there. It's actually my favorite example. You can dial it up/down very well to push at people's intuitions. On one extreme is where you're actually going to die if you don't cut the rope. You can dial this down to just some risk of dying. You can dial it down further to just some risk of harm (maybe it's cutting off circulation to your foot, and you might lose your foot.... or maybe it's just threatening to give you rope burn; are you justified in cutting the rope then?). This is a good example that poses some tough questions, but yeah, almost none of your other examples work at all.
ControlsFreak
3d ago
With that in mind, what does it mean for the price to be "wrong"? What's supposed to happen if the "efficient market" doesn't get the price "right"?
That depends significantly on the reasoning why one thinks that may be the case. The unfortunate part is that it's about the same level of impossibility to know for sure that you've gotten that reasoning right as it is to know for sure what the underlying value of future cash flows is.
That said, if we are allowed to handwave a bit, it has been suggested by far greater financial minds than I (e.g., Matt Levine), that some stocks are sometimes priced above the current value of their future cash flows because of memes (e.g., Gamestop). If that's the case, then one would expect that the price would follow the dynamics of memes, which may or may not be at all similar to the dynamics that one would expect the price to follow if one thought that it was primarily being priced by more 'traditional' concerns. What is "supposed" to happen depends significantly on things like, for example, how long one thinks that it will primarily follow the dynamics of memes or whether that component will eventually disappear and such.
Obviously, there are many directions that any company may end up taking, and there could be many different underlying reasons why a (slightly-modified) "reasonably-efficient market" doesn't get the price "right". For example, a company may be engaging in fraud, and perhaps the vast majority of market participants are unaware of this fraud. What has sometimes happened in the past in such cases is that the company runs out of money, suddenly declaring bankruptcy and surprising everyone by telling them that their shares are a claim on $0 worth of shareholder equity. In other cases, an external party discovers the fraud, and as that information spreads through the market, many people want to sell, but no one wants to buy, and the price quickly collapses. Other dynamics can occur if there are other factors.
ControlsFreak
6d ago
IIRC, the shortest I've seen was, "We have Roko's Basilisk at home"
ControlsFreak
6d ago
Suppose I have legacy VBA macros in Excel. Are there any that let me just directly use them? Last I had looked at LibreOffice, I think, and I was going to have to do some significant rewriting.
ControlsFreak
8d ago
Allow me to clarify. I'm aware of some of the weaknesses in the models that are used purely for the physical systems, but I actually don't think they're that bad. They're actually pretty decent, AFAICT, where "pretty decent" is a sort of term of art to describe situations where you have an okay sense for the scale of your error analysis and can sort of understand when it might be a problem or not a problem.
Sigh. I really don't know that I have a good way of explaining this intuition. Maybe a story. Long ago, I paid a lot more attention to fairly straightforward aircraft control topics at conferences. One question that came up surprisingly often was, "...would you fly on an airplane that is being controlled by your controller?" There is just a qualitative sense that you develop for how bad the badness is. How the error is likely to be structured, how likely genuinely destructive failure modes are, magnitudes of expected error under various noise distributions, etc.
With that, I guess I'll just say again that I don't think the climate models are "that bad". They're not garbage. There is error, we know some of the sources of error (you mention a couple), and we have an intuitive sense for about how big it may be.
Conversely, what I have significant theoretical disputes with is specifically trying to model economic systems coupled with climate systems. We basically can't even get off the ground, theoretically speaking. The timescales are the wrong way round. When you talk about the apparent fastness of climate dynamics, they are still figuratively glacial in comparison to the dynamics of economic systems (and I sometimes tack on political systems, because it's amazing how many people try to make completely whacko claims about these, too). You did great to realize that attempts like using a static damage function and then proceeding with a simple amortization are bonkers. My point is that the underlying theoretical reason why they're bonkers is because that's just not how one does anything with timescale-separated coupled dynamical systems.
ControlsFreak
9d ago
Have you actually examined the details of Nordhaus’ models?
Oh boy have I. I went through it extensively way back in the day, when we were at the old old old place. That experience was part of my coming to the conclusion that the entire endeavor is simply an impossible task. You mention some of the problems; there are others. I agree that his Nobel was a scientific travesty and sad state of affairs.
If you want to understand climate change and why it’s so catastrophic, you have to model and understand the complex interactive feedbacks and it’s abundantly clear that he doesn’t.
I would echo this, but with one very minor modification:
If you want to understand climate change and its interaction with economics, you have to model and understand the complex interactive feedbacks and it’s abundantly clear that he doesn’t.
You simply state that it's "catastropic". Whereas I think it's pretty much impossible to actually model the complex interactive feedbacks... especially when it comes to their intersection with political/economic systems.
But now due to human activity, the climate is changing faster than the response to it.
One of the problems with the whole sort of analysis you do in this paragraph is that everyone does timescale-separated coupled dynamical systems backwards in the case of political/economic-climate coupling. As I alluded to in my comment, the dynamics of political/economic systems are fast, much much much faster than the dynamics of climate, even with human activity. If one spends time with the theory of such coupled systems (the canonical text being Khalil's book), which I have done extensively for non-climate-related professional reasons and prior to engaging with any economic-climate models, then one understands the proper way to go about analyzing such systems. And, well, nobody does it the proper way. Why not? In my view, it's because they can't. It's impossible. Rather than the problem lying with human psychology, the problem is that the math doesn't math that way.
ControlsFreak
10d ago
Every so often, I see a number that strikes me in a particular way. More than once, the way that it strikes me has been in comparison to climate change damage estimates. Yes, yes, there are many many different estimates out there, and they're even presented in different terms, too. Some are in percentage of GDP/GWP; others are dollar figures. One of the numbers that has stuck in my brain, thanks to David Friedman back at the old old old place, comes from one of the early world leaders in trying to produce such estimates, Nobel-winning William Nordhaus. It would take epsilon more effort to find one of his old old old comments at the old old old place, so I just found an example from his substack.
Nordhaus’s final and most important point was based on his own research.
My research shows that there are indeed substantial net benefits from acting now rather than waiting fifty years. A look at Table 5-1 in my study A Question of Balance (2008) shows that the cost of waiting fifty years to begin reducing CO2 emissions is $2.3 trillion in 2005 prices. If we bring that number to today’s economy and prices, the loss from waiting is $4.1 trillion. Wars have been started over smaller sums.
What he does not mention is that his $4.1 trillion is a cost spread over the entire globe and an extended period of time. I initially assumed his calculations of cost were for the rest of the century, making his $4.1 trillion total about $48 billion a year, but in A Question of Balance he appears to be summing over the next 250 years which reduces the annual cost to $16 billion.
It's a quote from Nordhaus' 2012 NYT opinion piece, citing his 2008 book, so yeah, the estimate is quite old. There are many many other estimates out there since then, but this one stuck in my brain. I think he was trying to get it to stick in your brain. "Wars have been started over smaller sums," is meant to do that. It worked.
This morning, Tyler Cowen posted How Much Has Shale Gas Saved U.S. Consumers? It's just quoting an NBER working paper. I'll just reproduce the whole quote, so there's no need to click through:
It may seem like a distant memory now, but as of the mid-2000s, U.S. natural gas production had been flat for a decade, and the U.S. was importing liquefied natural gas (LNG), with plans to import much more. Then shale gas happened. Advances in hydraulic fracturing and horizontal drilling caused U.S. natural gas production to increase significantly, and the U.S. went from being a net importer of natural gas to being the world’s largest exporter. This paper calculates how much shale gas has saved U.S. natural gas consumers. Using price differences between the United States, Europe and Japan, we calculate that U.S. natural gas consumers have saved $3.1-$4.3 trillion between 2007 and 2025, equivalent to $164-$227 billion annually. Access to low-price U.S. natural gas has been particularly valuable during major supply shocks such as the war in Ukraine, and the benefits of shale gas have been experienced broadly across sectors and states.
It's not a direct analog, but that number, though. It's in my brain. $4.1T is right in that range of $3.1-4.3T. That's a swing in one country over less than 20 years, not 250 years. The dynamics of economic systems can move fast, much faster than climate change. But how big of a swing does that 'feel like'? Sure, life would have been more awful in a variety of ways in the counterfactual without the shale revolution. But, like, cataclysmically bad?! End of the world bad? I kind of doubt it.
I don't really like to focus too much on any particular estimate. There are higher ones; there are lower ones. I actually think the entire endeavor of estimating economic impacts of climate change is probably impossible, but we're stuck in a world where we have the various estimates we have and they matter to people. But I never underestimate how difficult the scale of numbers is to folks, so I appreciate when I occasionally see numbers of roughly similar scale in different contexts.
Ok, so I'm trying to follow. The clock we're analyzing has all of the access it needs in order to do y', which is what it's going to do, and which we've observed is suboptimal. But then, I guess, we're like, hypothesizing that we could conjure up some faster-than-light travel for something, waves hands, to this clock. And that, somehow, waves hands, I guess if we, like, change the design of the clock or something, which we can't do, somehow, waves hands, could end up in it doing action y instead of y'.
Like, what is the problem here? What is the space of possible actions? What am I trying to solve for?
ControlsFreak
24d ago
you don't have access
I don't follow. Every part of you that is necessary to follow the clockwork has appropriate access to the mechanisms of the clock, at least to the extent that is necessary for it to be able to follow the clockwork. If there were some part of you that didn't have such access, then it wouldn't be able to follow the clockwork, and we would reach a contradiction.
Like, maybe try to explain how this works directly on the example of analyzing an actual clock, with determined suboptimal action y' and a hypothesized optimal action y. What doesn't have access to what?
ControlsFreak
24d ago
I think I have now concluded my argument that there is no contradiction, once one tries to explain how the contradiction is supposed to work.
ControlsFreak
24d ago
I don't see a contradiction at all. This proposed unmoved mover is already clearly an exception to the general rule of requiring a prior mover, apparently preferring avoiding infinite regress over having an exception. It is simply not moved by some prior cause. That's kind of it? I think you'll have to be more explicit about how you find a contradiction.
ControlsFreak
24d ago
What's confusing is that I'm missing an argument. Some sort of, "Here are some premises, and here's a conclusion," sort of thing.
ControlsFreak
24d ago
I'm not quite seeing an argument yet. Go on?
I mean, kinda no? That's where the Wolpert/Benford critique comes in. You can't formalize the problem in terms of game theory without adding additional assumptions. If your additional assumptions to formalize it are, "It's actually a clock, and there's no feasible action set with cardinality greater than one," then sure, you have a suitable formalization... but it's kinda not game theory. If you want to back away from that being your additional assumption, it's kinda still on you to state other formal additional assumptions that make it a well-posed game.
EDIT: Perhaps another way of describing it would be as follows. Suppose one is just analyzing a clock. We'll discretize time for now just to make it simple. Say that we observe from our analysis that in the transition from time t_1 to t_2, the clock will become one second slow compared to some 'objective' time (handwave any difficulties here). We could observe that this is, in some sense, suboptimal, sure.
Now, does it make sense to say something like, "What if we just call this suboptimal action y' and hypothesize an alternative action y that doesn't result in being one second slow?" Would it make sense to say that we have constructed a decision theory problem? Note that we're not specifying anything about any sort of real policy space or anything; it's not like we're saying, "Here is the policy space of possible mechanisms that a non-clock can choose from to design the clock."1 We just have a clock.
Suppose we say that there is some being, Omega, who will accurately predict that said clock will take action y' and become one second slow, and then put some quantity of money in front of the clock. Suppose we say, "Well, imagine the clock took hypothetical action y, which it can't do, then imagine that Omega would put a different quantity of money in front of the clock in that case." Does this become a game theory problem? If so, what am I supposed to solve for? What is the space of possible solutions?
1 - This is perhaps related to my comment about what Yud did to the prisoner's dilemma problem. He created some different policy space about source codes.
ControlsFreak
24d ago
I think this is non-responsive to my comment. Isn't god himself a "mover" in classical theology?
ControlsFreak
25d ago
Jesus moves and changes yet he's the god that is not supposed to do either of those things
I already don't really follow. I thought the second word of "unmoved mover" was "mover". I didn't think classical theology posited an unmoved unmover.
ControlsFreak
25d ago
In a clockwork universe, is there such a thing as decision theory, or a subset thereof known as game theory? It would seem to me that, sure, one could have a mathematical theory of optimization, extremal values, or even min-max theory, but it would not seem to me that one could view any such results as being prescriptive - I.e., "If you are trying to accomplish X, you should choose Y." Instead, it would simply observe, "You might by chance (or deterministic integration of physical differential equations or whatever) take action Y or Y', and it turns out that we can compute that Y is optimal for purpose X, while Y' is suboptimal."
That is, if one is an adherent to this conception of a clockwork universe, I think the way they would state their position on Newcomb's problem would be something more like, "You will either 1-box or 2-box, based on the movements of the clock. We can also compute from axioms regarding the clock's movements that 1-boxers will possess more money," and less like, "You're in this hypothetical situation where you need to think about the rational way to proceed optimally, and here is why you should choose to act in the following way." I think if such proponents presented their perspective in this way, it would be less amenable to criticism that their problem is ill-posed as a decision/game theory problem.
The axioms of decision/game theory seem to conflict with the axioms that seem to appear here. I guess one way to put words to that would be that one has a feasible action set within the underlying dynamical system that has cardinality greater than one. Perhaps another way to put it is that it does not seem to me that decision/game theory is applicable to clocks. The feasible action set of clocks has cardinality one. One does not ask how a clock should choose among non-identical actions, though one may observe whether a clock's deterministic actions are/are not optimal according to some metric.
Taking this alternative position would, I think, sidestep the criticism I relayed above from Wolpert/Benford, as what they were fundamentally trying to do was to formalize the problem within decision/game theory, where players have feasible action sets with cardinality larger than one. They observed that if you do this, you run into contradictions without further specification. But it would seem like, sure, if you give up on that, give up on saying that it has anything to do with decision/game theory, that it's more like just making an observation about clocks and optimality/suboptimality, then I think you do avoid the critique.
ControlsFreak
26d ago
My sense from the text is probably annihilationism.
You probably have to go more for AI atheists to find a god that makes future contingents like sea battles or box picking necessary and then also sets up a basilisk to torture you forever for the future contingents that it retroactively would have made necessary.
ControlsFreak
27d ago
My sense tracks with that of @MathWizard. If you add some particular assumptions about the form of the problem, you can code it up, and likely, for a wide range of parameters, 1-boxing is higher EV.
I think the criticism of Wolpert/Benford is also similar in type. (Again, not really having spent sufficient time with it.) That is, they construct two possible interpretations. Either of them, you could just sit down and code. It may even be the case that for a wide range of parameters, EV still points to 1-boxing for both versions. However, my understanding of their claim is that those two codes will be very different. Even the strategy spaces are fundamentally different in their claim. And for a similarly wide range of parameters, the joint distributions will be contradictory. The point is not that the sign may be the same for this particular ratio of prizes; it's that there are just multiple contradictory ways to construct it.
Of course, someone could take the time and search out what ratio of prizes in the respective boxes produces maximum tension between the two interpretations, so that rather than having the two EV calcs mostly pointing in the same direction, we could maximize how often they conflict. That's kind of not the point of the critique, but I suppose it could be done if one found it necessary to really grok the difference between a well-posed and ill-posed problem. Though, like you put it, I probably can't be arsed to do it.
That said, I am almost motivated enough to try it (but it would probably have to wait a few weeks, and then, I'll probably be bored with it). I certainly don't know that we can for sure find parameters where the two possible games differ in terms of sign. If this problem was actually relevant to my research interests, I would absolutely just do it, because it's one where I have a vague sense of, "Wouldn't it have to be amazingly coincidental if the values were different, but the signs were always the same?" And when I sniff at the possibility that there could be an amazing coincidence like that, it's usually an indicator of a really interesting theoretical opportunity.
ControlsFreak
27d ago
Wolpert and Benford argue that the problem is ill-posed for almost any error rate, so it's not clear that stuffing in a particular number actually helps resolve the problem. I haven't spent all that much time with this problem yet, so I'm not going to commit to saying that I think they're right about this, but it jives with my intuition.
Generally speaking, in order to have a well-posed game, one must be very formal and precise in many details. Particularly things like order of operations, allowable policy spaces, information sets, and details around estimators. I've become more annoyed by estimators in various problems over time, even apart from the relatively minimal thinking I've done on Newcomb's problem. One of the greatest sources of my criticisms in reviews of submitted papers (or even when my collaborators come to me with a problem set-up and/or proposed solution) revolves around not taking sufficient care around estimators.
I do think that Wolpert/Benford at least suffice in arguing that there are at least two possible formalizations that are sufficiently well-posed. I think it's probably on someone else to either bite the bullet and say they are clearly choosing one form or the other... or to provide a sufficient alternative formalization that makes the details more clear.
Aside on Yudkowsky, relevant for the discussion below and my thoughts generally on these sorts of problems. I wouldn't be surprised if he has/had something in mind like what he did to the prisoners' dilemma problem, with the business about source codes and such. There could be a way to try to resolve Newcomb's problem in a similar fashion, but my perspective is that it would still be proposing a very specific formalization... and one that is not at all just a clear instantiation of the initial problem statement. I might go so far as to say that in the prisoners' dilemma case, he just proposed a different problem, with different policy spaces. Interesting in its own right, sure. Probably correct for that particular formalization of that particular version of the problem, sure. But also kind of just a different problem. In general, even minor tweaks to these aspects of the formulation can result in different games.
Similarly for Newcomb's problem, unless one takes the step of clearly laying out in a formal way exactly what they're going to specify for the domain of the problem (and then, I guess, argue that this is like, 'the one true interpretation of the original problem' or something), then I'm probably going to lean toward just thinking that the original problem is so informally stated as to be ill-posed.
ControlsFreak
1mo ago
Against Talking About Anthropics/Possible Worlds/etc in the Sleeping Beauty Problem
I get it. Anthropics is an interesting topic. Possible worlds has a long and rich philosophical history. I can get why people might want to expose more people to that stuff, kinda squint at the sleeping beauty problem, then think that it's close enough to spread the gospel.
But that's confusing people.
It's confusing them on what is otherwise a very simple math problem.
For those who haven't seen my last entry, I made some minor modifications, primarily adding a second person, so we have both Alice and Bob undergoing simultaneous experiments. The simplest version is that they each undergo approximately the same experiment, with the same coin, but opposite (the implications of heads for one person are like the implications of tails for the other). I also had some computer communication between them for some instructive purposes, but that's not even necessary here.1
Let's follow Alice and Bob a bit further. Suppose after their one/two awakenings, they're put back to sleep, memory again wiped. They're both finally awoken on Wednesday. "No more questions," the doctors say. "We took the liberty of interpreting your answers as wagers. We have your home address. We'll compute your payout and mail you a check with your results, revealing to you how the coin actually came out, how you answered the questions (because they won't remember), and what your payout is. Expect it to take 4-8 weeks."
Alice and Bob leave their respective rooms. They run into each other in the lobby.
Wait
Can Alice and Bob run into each other in the lobby? Aren't they, like, in different possible worlds or something? No, silly. That's confusing people. They're in the same world. They've been in the same hallway all along, separated by only a paper-thin wall.
Ok, so they run into each other in the lobby. They hit it off, decide to go out to a pub and grab a pint together. Naturally, the conversation turns to the strange experiment they each went through. Neither one is going to know how the coin flip actually went or what subsequently happened for another 4-8 weeks.
They begin to debate. How should they best guess what their results might have been? What if they'd like to wager against one another about the results? Should they have significantly different estimates of what they're going to see in their results? Should Alice think that there's a 1/3 chance that they're going to learn that it was heads, while Bob should think that there's a 1/3 chance that they're going to learn that it was tails? Did they truly "update" their probabilities during the course of the experiment?
No. Of course not. If either of them thought that, you could take their money. They should both think that it was 1/2 either heads or tails. This is because they didn't "update" some probability estimate. They didn't enter some weird different possible worlds, where never the physical Alice and Bob could ever meet again.
Instead, Alice and Bob are both capable of having a perfectly reasonable conversation. "Yeah, of course I think the probability of the coin flip was 1/2. It's just because of the weird observation function of the experiment that I computed that there was a different probability for what I was likely to observe." "Yeah, me too, but my observation function was the opposite, so I computed that I was likely to observe the opposite. But obviously, at the same time, the probability of the coin flip was 1/2."
They're just different probabilities with different meanings. You can just compute them from the observation functions.
1 - That time, I was trying to get people to figure out that they could have one individual's brain retaining multiple different probabilities, with multiple different meanings. I guess this time, I'll just try having multiple different minds meeting.
ControlsFreak
1mo ago
Perhaps part of it is that married women who changed their name want to vote too.
As someone whose wife came from a foreign location where women don't tend to change their names, and can thus attest to a significantly higher-than-normal level of grief over the wife changing her name, getting US documentation that would be sufficient for voting is probably the easiest part of a married woman changing her name.
What sort of engineer? What sort of regulation/code? I want to profit.
More options
Context Copy link