Froolow

Karma: 250

Froolow Apr 30, 2014, 3:16 PM
0 points
in reply to: brazil84’s comment on: The Cryonics Strategy Space
While on the one hand I completely agree with you given your starting premises, I don’t necessarily think we’re in quite the zero information situation you describe. For example, it is pretty well accepted (even amongst people who don’t think cryo will work) that simply freezing yourself without cryopreservant lowers your chance of revivification. This is a pretty important consensus since cryopreservant is highly toxic, but we extrapolate from current trends and conclude, “Curing poisoning is probably an easier task than reconstructing information destroyed by entropy, so I should adopt the ‘cryopreservant’ branch of strategy-space”. This indicates we don’t really have no information about the correct cryo strategy; though I totally accept your weaker claim that I seem to demand much MORE information than we can reasonably be expected to possess.

I think we’re in a situation more like a friend ringing up and says, “We’re going to play Ticket to Ride tonight; it’s like Monopoly only better”. We don’t have enough information to decide whether we want to be the top hat or the battleship (which is a meaningless question anyway since the answer is always ‘top hat’), but we might have enough information to begin to say, “On my first turn I will study the layout of the board carefully (rather than act quickly)” and “I will attempt to remain on good terms with the other players insofar as they can hurt me and I cannot overwhelmingly hurt them” or even “It is unlikely this game will involve serious roleplay. I will not put on my robe and wizard hat”. None of these are enough to guarentee a win, but neither are they trivial realisations; I think it is reasonable to believe probability theory, human nature and my own utility function will not change dramatically in the time it takes me to be revivified, so basing strategy on these characteristics seems worthwhile.

Froolow Apr 26, 2014, 10:05 AM
0 points
in reply to: CronoDAS’s comment on: The Cryonics Strategy Space
You could try signalling that unless they trade with you you’ll put them at a disadvantage. Consider—“Player 1, both you and Player 2 want this property. This property for one of your properties is a fair swap that benefits both of us, but if you turn me down I’ll trade it to Player 2 for their best offer, which benefits Player 2 a lot and me only a little. Player 2, if you don’t give me even the small amount I ask for, I’ll randomly give it to some other player.” If you signal credibly (ie you actually do it if someone calls your bluff) then Player 1 should make the trade provided he values your property more than you knocking yourself out of the game (ie the trade really DOES have to be a fair deal—you can’t just use this to up your bargaining ante).

Part of your argument could be the (truthful) observation that the losers in a trade aren’t the people who made the less valuable trade, but the people who didn’t trade at all—if you play at a very conservative table it might be in your interest to trade at a disadvantage and exploit the increased variance a monopoly gives you.

The major downside here is that many people don’t play Monopoly to win, and strategies that optimise for winning using game theory and the like are seen as ‘unsporting’. Sometimes you just need to accept that to keep your family happy you have to play Monopoly and get bored.

Froolow Apr 25, 2014, 2:06 PM
1 point
in reply to: Izeinwinter’s comment on: The Cryonics Strategy Space
I’m not sure I completely agree with you, but I’d argue that is exactly the sort of discussion which I am surprised is not already happening. Consider:
- I should not make myself an appealing target for resurrection, because I am likely to receive the procedure in the most ‘pre-alpha’ form
versus
- I should make myself the most appealing target for ressurection possible; history shows that if a procedure is expensive or difficult (like going to the moon) it is usually only done infrequently until technology catches up with ambition. The longer I am frozen, the more chance something happens to catastrophically prevent my resurrection, so I desire to be revived in the first wave.
Alternatively
- Future society is likely to punish (or refuse to revive) those who were evil in this life, so I should only adopt strategies which reflect well on me
versus
- Future society is likely to reprogramme people who were evil in this life before reviving them, so I should maximise my chance of making it to future society by any means necessary; it won’t affect my chances of revivification because almost every current human will need reprogramming before revival.
While it happens my probability distribution over what future society looks like is a lot closer to yours than what you might infer from the main body of the post, your belief about future society is hugely important in determining your optimal freezing strategy. This is why I say I am surprised there is not more discussion already happening; cryonics correlates well with people who have carefully considered the question of what future society is likely to look like, but it appears many people have not then made a link between the two sets of beliefs.

Froolow Apr 25, 2014, 1:56 PM
−1 points
in reply to: seez’s comment on: The Cryonics Strategy Space
Personally, it seems like a pretty rational decision to me (excluding autopsy problems, which I talk about somewhere else). The reason I advise against it is because I don’t believe anyone could possibly know their utility function—and life expectancy—well enough to make a sensible decision about when the right time was to begin the process. This is true even if you exclude the fact that there are good reasons to think that many people do not approach death rationally, and if you consider that an ostentatious decapitation would likely be distressing for those left behind (insofar as you care about the utility of people after your death).

But in a purely hypothetical case—where there was a bomb in my heart that was going to go off in ten seconds and I happened to be standing next to a big vat of cryopreservant—I would highly recommend freezing yourself before dying naturally.

Froolow Apr 25, 2014, 1:47 PM
2 points
in reply to: V_V’s comment on: The Cryonics Strategy Space
I agree the fact that current cryonics practice is not optimised for revival is extremely strong Baysian evidence (for me at least) that most cryonicists on these forums are considerably more likely to be signalling than rationally trying to live forever. I would add into that the well known problem of ‘cryo-crastinating’, which is hard to explain if pro-cryo individuals are highly rational life-year maximisers, but extremely easy to explain if people are willing to send a ‘pro-cryo’ signal when it is free, but not when it is expensive.

On the other hand, I am convinced at least some cryonics advocates will find a discussion on cryonics strategy genuinely useful and important, and since I was reading about cryonics anyway I thought I would fill what I considered to be a gap in the debate landscape.

Froolow Apr 25, 2014, 1:38 PM
5 points
in reply to: seez’s comment on: The Cryonics Strategy Space
Unfortunately my friends would probably see winning too often as a good reason to collude against me. Although collusion would lower the average length of a game, it would probably raise the chance any individual friend wanted to play with me (because they would be winning more often, on average). Although I agree with you that that’s a strategy I hadn’t considered, which is quite an oversight given the content of the post!

Khoth has correctly identified that surely the best strategy is to convince my friends to play a similar but superior game, although this isn’t always possible. For example with the horse-traders I try to play Catan and with the roll-and-movers I play Pirates. Unfortunately if there are too many of both groups then the only thing they can compromise on is Monopoly, and I don’t have the persuasive skills to overcome the inertia.

However the fact there are a whole bunch of superior games to Monopoly sort of breaks the analogy I was driving at so I left it out of the main body of the post.

Froolow Apr 25, 2014, 1:20 PM
2 points
in reply to: Lalartu’s comment on: The Cryonics Strategy Space
You are completely correct about Alcor’s FAQ:

While some communities have enacted legislation allowing suicide with the assistance of a physician, any such case almost certainly would be followed by an autopsy which would include dissection of the brain. For these reasons, and to protect ourselves from any accusation of conflict of interest, Alcor has a strict policy against advising any member to end life prematurely.

However this identifies exactly the point I was making; a rational discussion could be had about the risk of autopsy destroying the brain versus the benefit of being able to very tightly control your freezing. To expand on this; autopsy in the case of suicide is mandatory in the US to determine cause of death. I strongly suspect that if the average coroner comes across a headless frozen body full of toxic cryoprotectant they can determine the cause of death without destroying the brain, and if they can determine the cause of death witout destroying the brain they might choose to respect the implicit wishes of the deceased and explicit wishes of surviving relatives not to cut into the brain tissue. By contrast, freezing the brain ‘in use’ might increase the chance of survival. If an ostentatious suicide raises your chance of an autopsy by less than it raises your chances of revivification, it is a plausible strategy.

I’d add that Alcor has a very, very strong reason to advise members against suicide which members themselves do not have; Alcor can get sued for that sort of behaviour.

With respect to your other point—that these strategies are not novel—I can only agree that I would be surprised if I were the only person to have thought of them, but I did not come across any serious discussion of them even after some fairly comitted googling. If somebody looking for discussion of these strategies can’t find them, the odds of someone interested in cryonics but paradigm-bound to ignore the possibility of other strategies will have even more difficulty; it is for those people this article is written.

Froolow Apr 24, 2014, 11:32 AM
0 points
in reply to: odriew’s comment on: How long will Alcor be around?
One possible mechanism would be a general social shift towards more cryogenics meaning cryo voters became an important voting block. Since most rational cryo-voters can be expected to be more-or-less single issue with respect to cryonics (almost nothing will increase your individual expected utility for a given level of money more than increasing your chance of being revivified), politicians will begin to face great pressure to appease this demographic. You’ll see that this is different to the situation you describe for at least three reasons:
- On those issues where the individual utility gain is greatest, the population is smallest (cures for very rare genetic conditions which are unaffordable to the average person and yet not subsidised by the government). This is probably because it is not in the interests of politicians to use political capital on a very small sub-section of the population.
- On those issues where individual utility gain is small and populations are large, the individuals concerned are unlikely to be single-issue. For example public health measures undoubtedly raise my lifetime utility, but do they do so more than public education, public art or nebulous concepts like ‘freedom’? Hard to say
- On those issues where individual utility gain is large and populations are large, those populations are almost inevitably located in areas where US politicians have no incentive to help them. For example, campaigning to end malaria would be both massively important and affect a huge number of people, but those people would not be US voters.
If this social shift occurs, politicians may be incentivised to offer a ‘government guarentee’ to all frozen corpsicles, in the same way all mortage lenders are government-backed or banks are unable to go bust in an uncontrolled way (assets up to a certain value are protected). So it wouldn’t so much be a ‘not-death’ right (because all three groups I describe above would still fail to be protected from death), but I was using it as a shorthand for the slightly more complex scenario I describe here.

I don’t know how likely I think this scenario is, but I think if it is going to happen, it will happen before a post-scarcity society. In the interests of being charitable to the cryogenics companies, I think it is fair to point out that this is a mechanism that could greatly improve their chance of being revivified without any technological innovation.

Froolow Apr 18, 2014, 5:44 PM
4 points
in reply to: Eliezer Yudkowsky’s comment on: How long will Alcor be around?
I understand the concern about unpacking bias, and read about a related experiment also by Kahneman (I think) who elicited a higher probability when he asked experts to estimate the likelihood a specific scenario (deflation of the rouble leads to a Soviet invasion of Germany and nuclear war) than a general scenario (nuclear war). So I would be cautious of handling an equation with multiple, obviously overlapping terms. I’ll update the original post when I’m back at a computer to include a health warning in the first paragraph.

I don’t think I fully understand the criticism of this piece though; are you saying the modelling approach is incoherent or simply cautioning people not to just plug it into the cryo-Drake equation without considering the unpacking bias?

Froolow Apr 18, 2014, 8:55 AM
1 point
in reply to: lincolnquirk’s comment on: How long will Alcor be around?
Other relevant differences might be that humans are never allowed to just ‘die’ of making bad financial decisions in countries like America—if humans make really wild spending decisions the state will at least feed and house them.

Perhaps charities would be a better reference class? If anyone can find any data I’ll happily rerun the analysis, but ‘age charities’ will give you charities concerned with age and ‘life expectancy charities’ will give you charities concerned with life expectancy; it could be a bit of a slog.

Froolow Apr 18, 2014, 8:40 AM
1 point
in reply to: drethelin’s comment on: How long will Alcor be around?
Be careful about reading too much into that—“Large enterprises, those with 250 or greater employment, accounted for only 0.4 per cent of all enterprises.” according to the ONS. You’d expect to see 89.4% small companies by chance alone, although I concede that if a company is around for 100 years you might expect it to grow into a large company by inertia alone.

With respect to your other point, you are absolutely right—I wanted to show my working here to indicate how badly wrong back-of-the-envelope calculations can go in situations like this.

Froolow Apr 18, 2014, 8:35 AM
2 points
in reply to: lavalamp’s comment on: How long will Alcor be around?
Although in reality it makes a big difference, in my model it does not—my model varies only the size of the company, since that’s all I could find good data on. I found another source saying that the age of a company was about 30% more important in predicting its survival than its size, but because it was a complicated regression I was unable to exclude terms that had absolutely nothing to do with cryonics.

It is probable that you should shade the probability of Alcor surviving up and the probability of KryoRus surviving down to account for this.

Froolow Apr 18, 2014, 8:30 AM
5 points
in reply to: RolfAndreassen’s comment on: How long will Alcor be around?
Hi RolfAndreassen,

I’m impressed you spotted that so quickly, because it was non-obvious to me. Nevertheless, I did spot the problem you are describing and attempted to correct for it in the second graph by considering only companies which went into liquidation, using a proper academic source.

Froolow Apr 17, 2014, 11:01 AM
2 points
in reply to: Eugine_Nier’s comment on: Stranger Than History
This seems an unfair response to me—TheAncientGeek offers a standard argument pro-AA while admitting they haven’t studied the issue in detail. You attack the response on grounds that an AA supporter could rebut without ever contradicting themselves (i.e. “It isn’t collective justice, it compensates for individual inequality of opportunity (unless, say, you choose to define a progressive income tax as ‘collective justice’ in which case I do support collective justice)”, “It applies to certain minorities and not others because of the size of the disopportunity facing them (discriminatory social structures don’t distinguish between recent immigrants and descendants of slaves, but they do appear to discriminate between black African and white Irish)” and “It isn’t economically inefficient, and might even be economically efficient”).

The next paragraph contains an argument for AA which I support which I think proves there is at least one rational argument for AA. If it is important to you, I can also defend my position to prove to you it is not obviously wrong (although I hope the argument alone will be enough). If there is a rational argument in favour of AA, then there must be at least one utility function that makes supporting AA rational (in the same way that a utility function which really REALLY values ants might rationally choose to try to ban glasses so children can’t use them to burn ants). I don’t agree with TheAncientGeek’s starting premise that we should therefore suppress research into race, but I think it is important you don’t base your conclusion on a faulty premise (“AA is obviously wrong”).

This 2005 paper published in The Journal of Economic Education gives the result of an experiment where participants were randomly assigned a colour (‘green’ or ‘purple’) and given the following information (I’m paraphrasing badly to ensure I remain brief, please consult the paper for the actual protocol): “You are allowed to get education, which costs £1. You then take a (simulated) test where your score is randomly picked from 1 to 100, but if you bought education the score will have a small bias towards the higher end. ‘Employers’ (other participants) will then choose whether to ‘employ’ you. They only know your colour and your test score. If they employ you, you get £5. If they don’t, you get £1. If an employer picks an individual with education, the employer gets £10, otherwise they get nothing.” I presume the experiment was then iterated an unknown number of times to prevent gaming, but I can’t find that in the paper. Clearly, the socially optimal outcome is that everybody gets education and the employers employ everybody. However, individuals can earn the full £10 rather than a net £9 by gambling on the employers being over-generous and picking them even though they didn’t get education.

By chance, the ‘purples’ happened to be under-educated in the first round, which meant some purples who got an education decided not to waste the money next round. This therefore compounded the effect, to the point where new purples realised there was no point in investing in education, so even some free-riding greens couldn’t prevent employers betting on greens (even if the green score was lower than the purple score). If the society in the experiment were allowed to implement AA they would; it would be hugely more economically efficient to remove the pro-green bias and both encourage purples back into education and force greens to keep up their initial levels of education and not ‘free ride’. The experimental confirmation that AA can be economically efficient is reason enough to support such policies, but I think they would be more effective in the real world compared to the experimental world; for example, two contradictory opinions are likely to lead to more economic progress than two homogenous opinions, and this cultural bonus is not modelled in the original experiment.

Froolow Apr 14, 2014, 3:22 PM
14 points
on: [Requesting advice] Problems with optimizing my life as a high school student
In the particular case of the first problem there may be a shortcut that is worthwhile exploring. As I see it, your problem is that you would like to know how much leisure time to allocate to improve your productiveness (including the possibility of zero leisure time). The ‘improving of productiveness’ is the important goal to you, not the philosophical distinctions regarding the optimal work/life balance. Since productivity is something approximately measurable, you yourself can optimise over this domain.

With that in mind, you can perform an experiment on yourself. Start by allocating an amount of leisure time you think is excessive, but not wildly so. You want to pick a number that means you will be completely relaxed when you attempt to perform productive work, such that your productivity is ‘100%’ (however you want to define that). When I was at university I needed between one and two hours a day, now I work a full-time job I need closer to three. I’d suggest based on my experience alone that two hours a day would be a good starting point. Force yourself to have this much leisure (but to optimise substantially, train yourself that activities like exercise, cooking and meditation are pleasurable). If you find yourself worrying about not benefitting the future, think to yourself, “I am currently engaged in an experiment on myself, the results of which could make me substantially more productive for the rest of my life. It is highly unlikely that the insights of a marginal two hours’ work will benefit the future more than the insights from this experiment.”

After the end of your first week of this, reflect on whether you think you could reduce the number of leisure hours you spent and maintain your productivity. In particular, you should reflect on whether you can achieve the same level of fun in a shorter space of time, or whether you can decrease your marginal fun without decreasing your productivity. For example, would a ten minute break each hour be more refreshing than a two-hour game of Civ? Reduce your leisure hours by a small fraction of their total value; maybe schedule ten minutes less leisure next week. Repeat. It is important you don’t decrease fun too quickly or too sharply; you need to have a slow-ish period of optimising your fun.

Eventually, you will come to the point where you cannot possibly decrease fun without cutting into productivity. Here you want to make the decreases in your scheduled leisure time much shorter, and try to track more closely the impact they have on your productivity, such that you can identify the point where a marginal minute would be better spent resting than working. Remember that productivity isn’t simply the ability to churn out mediocre code with few errors, but the possibility to have a ‘brainwave’ and capitalise on it. After all, Friendly AI only needs to be solved once! Personally, I think I would happily take an extra half-hour out of my day if it meant I could guarantee I would be working perfectly productively for the rest of that day, but if your cost (in time) is high for a marginal unit of productivity you might differ.

Froolow Apr 9, 2014, 2:12 PM
8 points
on: Siren worlds and the perils of over-optimised search
This puts me in mind of a thought experiment Yvain posted a while ago (I’m certain he’s not the original author, but I can’t for the life of me track it any further back than his LiveJournal):

“A man has a machine with a button on it. If you press the button, there is a one in five million chance that you will die immediately; otherwise, nothing happens. He offers you some money to press the button once. What do you do? Do you refuse to press it for any amount? If not, how much money would convince you to press the button?”

This is – I think – analogous to your ‘siren world’ thought experiment. Rather than pushing the button once for £X, every time you push the button the AI simulates a new future world and at any point you can stop and implement the future that looks best to you. You have a small probability of uncovering a siren world, which you will be forced to choose because it will appear almost perfect (although you may keep pressing the button after uncovering the siren world and uncover an even more deviously concealed siren, or even a utopia which is better than the original siren). How often do you simulate future worlds before forcing yourself to implement the best so far to maximize your expected utility?

Obviously the answer depends on how probable siren worlds are and how likely it is that the current world will be overtaken by a superior world on the next press (which is equivalent to a function where the probability of earning money on the next press is inversely related to how much money you already have). In fact, if the probability of a siren world is sufficiently low, it may be worthwhile to take the risk of generating worlds without constraints in case the AI can simulate a world substantially better than the best-optimised world changing only the 25 yes-no questions, even if we know that the 25 yes-no questions will produce a highly livable world.

Of course, if the AI can lie to you about whether a world is good or not (which seems likely) or can produce possible worlds in a non-random fashion, increasing the risk of generating a siren world (which also seems likely) then you should never push the button, because of the risk you would be unable to stop yourself implementing the siren world which – almost inevitably – be generated on the first try. If we can prove the best-possible utopia is better than the best-possible siren even given IC constraints (which seems unlikely) or that the AI we have is definitely Friendly (could happen, you never know… :p ) then we should push the button an infinite number of times. But excluding these edge cases, it seems likely the optimal decision will not be constrained in the way you describe, but more likely an unconstrained but non-exhaustive search – a finite number of pushes on our random-world button rather than an exhaustive search of a constrained possibility space.

Froolow Mar 27, 2014, 9:37 AM
0 points
in reply to: elharo’s comment on: Political Skills which Increase Income
Hi elharo,

Your criticism is absolutely correct; not all writers are novelist so even if novelists show the income variation I assert, that wouldn’t show up on the BLS statistics. I think that shows that my illustrative example is flawed, but I hope it doesn’t undermine the main conclusion too much.

Froolow Mar 27, 2014, 9:32 AM
0 points
in reply to: Eliezer Yudkowsky’s comment on: Political Skills which Increase Income
Ah yes, this makes a lot of sense and explains my earlier confusion; although it may still be true that there is a high variance in income between novelists, not all writers are novelists (for that matter, I suppose not all novelists are writers, at least as far as the BLS will bin them). I think that indicates my illustrative example is flawed, although I hope the wider point still stands.

Froolow Mar 26, 2014, 4:02 PM
3 points
on: Political Skills which Increase Income
These findings might also be useful for choosing between high-variance and low-variance careers, insofar as you are able to predict how much better at each of these skills you are than average.

For example, engineering is a field where most people earn a decent wage and some people earn a very decent (but not obscene) wage. I think the average salary of an engineering graduate is about $58,000, with the vast majority of that made up by people working in the $30,000-$50,000 band with a couple of hotshots pulling down six-digit salaries and almost nobody grinding out a subsistence wage. By contrast writing a book is a field where one or two people become megastar billionaires like J K Rowling or John Grisham and everybody else earns practically nothing. Pretty amazingly (I think it’s amazing) the Bureau of Labour Statistics reckons the average wage for a writer is almost exactly the same as the average wage of an engineer ($56,000), but it seems likely to me the median salary is much lower than in engineering (maybe something like $15,000-$25,000) with the mean heavily skewed by the handful of super-rich authors.

What this suggests is that you can – on a probabilistic basis – determine whether you are likely to end up at the top or bottom of the income distribution of your chosen profession. If you know your political knowledge and skills are weak, it would be a good idea to pick engineering over writing (assuming you are equally good at both) because you are more likely to end up earning $30,000 than $100,000. If you are excellent at office politics you are much more likely to form the sort of connections that give you a bestseller, and so writing might offer you the highest expected earnings even though the average salary in both professions is (nearly) identical.

This is complicated by the fact that ‘political knowledge and skills’ are not consistent between careers (the most insensitive politician is likely to be far more manipulative than the smoothest political operator in database administration), but I think it is probably possible to allow for this and still have more information about career choice than you did before.