This post suffers from lumping together orthogonal issues and conclusions from them. Let's consider individually the following claims:
A priori, from (8) we can conclude (9). But assuming the a priori improbable (7), (8) is a rational thing for X to conclude, and (9) doesn't automatically follow. So, at this level of analysis, in deciding whether X is overconfident, we must necessarily evaluate (7). In most cases, (7) is obviously implausible, but the post itself suggests one pattern for recognizing when it isn't:
...The modern world is sufficiently complicated so that no human n
Unknown reminds me that Multifoliaterose said this:
The modern world is sufficiently complicated so that no human no matter how talented can have good reason to believe himself or herself to be the most important person in human history without actually doing something which very visibly and decisively alters the fate of humanity. At present, anybody who holds such a belief is suffering from extreme delusions of grandeur.
This makes explicit something I thought I was going to have to tease out of multi, so my response would roughly go as follows:
Upvoted for being clever.
That's interesting. I downvoted it for being clever. It was a convoluted elaboration of a trivial technicality that only applies if you make the most convenient (for Eliezer) interpretation of multi's words. This kind of response may win someone a debating contest in high school but it certainly isn't what I would expect from someone well versed in the rationalism sequences, much less their author.
I don't pay all that much attention to what multi says (no offence intended to multi) but I pay close attention to what Eliezer does. I am overwhelmingly convinced of Eliezer's cleverness and brilliance as a rationalism theorist. Everything else, well, that's a lot more blurry.
Success is not assured. I'm not sure what's meant by confessing to being "ambitious". Is it like being "optimistic"? I suppose there are people who can say "I'm being optimistic" without being aware that they are instantiating Moore's Paradox but I am not one of them.
I also disclaim that I do not believe myself to be the protagonist, because the world is not a story, and does not have a plot.
give the SIAI staff an opportunity to respond to the points which I raise in the present post as well as my two posts titled Existential Risk and Public Relations and Other Existential Risks.
Indeed, given how busy everyone at SIAI has been with the Summit and the academic workshop following it, it is not surprising that there has not been much response from SIAI. I was only involved as an attendee of the Summit, and even I am only now able to find time to sit down and write something in response. At any rate, as a donor and former visiting fellow, I am only loosely affiliated with SIAI, and my comments here are solely my own, although my thoughts are certainly influenced by observations of the organization and conversation with those at SIAI. I don’t have the time/knowledge to address everything in your posts, but I wanted to say a couple of things.
I don’t disagree with you that SIAI has certain public relations problems. (Frankly, I doubt anyone at SIAI would disagree with that.) There is a lot of attention and discussion at SIAI about how to best spread knowledge about existential risks and to avoid sounding like a fringe/doomsday organization in doing so. It’s true that...
I'd like to vote this up as I agree with lots of the points raised, but I am not comfortable with the personal nature of this article. I'd much rather the bits personal to Eliezer be sent via email.
Probably some strange drama avoidance thing on my part. On the other hand I'm not sure Eliezer would have a problem writing a piece like this about someone else.
I've thought to myself that I have read one too many fantasy books as a kid, so the partying metaphor hits home.
I'd like to vote this up as I agree with lots of the points raised, but I am not comfortable with the personal nature of this article. I'd much rather the bits personal to Eliezer be sent via email.
I was conflicted about posting in the way that I did precisely for the reason that you describe, but after careful consideration decided that the benefits outweighed the costs, in part because Eliezer does not appear to be reading the private messages that I send him.
I'm inclined to think that Eliezer's clear confidence in his own very high intelligence and his apparent high estimation of his expected importance (not the dictionary-definition "expected", but rather, measured as an expected quantity the usual way) are not actually unwarranted, and only violate the social taboo against admitting to thinking highly of one's own intelligence and potential impact on the world, but I hope he does take away from this a greater sense of the importance of a "the customer is always right" attitude in managing his image as a public-ish figure. Obviously the customer is not always right, but sometimes you have to act like they are if you want to get/keep them as your customer... justified or not, there seems to be something about this whole endeavour (including but not limited to Eliezer's writings) that makes people think !!!CRAZY!!! and !!!DOOMSDAY CULT!!!, and even if is really they who are the crazy ones, they are nevertheless the people who populate this crazy world we're trying to fix, and the solution can't always just be "read the sequences until you're rational enough to see why this makes sense".
I realize it's a bala...
there seems to be something about this whole endeavour (including but not limited to Eliezer's writings) that makes people think !!!CRAZY!!! and !!!DOOMSDAY CULT!!!,
Yes, and it's called "pattern completion", the same effect that makes people think "Singularitarians believe that only people who believe in the Singularity will be saved".
Yeah, "people who think Singularitarians think that" is what I meant.
I've actually met exactly one something-like-a-Singularitarian who did think something-like-that — it was at one of the Bay Area meetups, so you may or may not have talked to him, but anyway, he was saying that only people who invent or otherwise contribute to the development of Singularity technology would "deserve" to actually benefit from a positive Singularity. He wasn't exactly saying he believed that the nonbelievers would be left to languish when cometh the Singularity, but he seemed to be saying that they should.
Also, I think he tried to convert me to Objectivism.
Do otherwise and YOU and YOUR LOVED ONES will suffer ETERNAL OBLIVION.
This one isn't right, and is a big difference between religion and threats like extinction-level asteroids or AI disasters: one can free-ride if that's one's practice in collective action problems.
This whole "outside view" methodology, where you insist on arguing from ignorance even where you have additional knowledge, is insane (outside of avoiding the specific biases such as planning fallacy induced by making additional detail available to your mind, where you indirectly benefit from basing your decision on ignorance).
In many cases outside view, and in particular reference class tennis, is a form of filtering the evidence, and thus "not technically" lying, a tool of anti-epistemology and dark arts, fit for deceiving yourself and others.
We all already know about this pattern match. Its reiteration is boring and detracts from the conversation.
A number of people have mentioned the seemingly-unimpeachable reputation of the Future of Humanity Institute without mentioning that its director, Nick Bostrom, fairly obviously has a high opinion of Eliezer (e.g., he invited him to contribute not one but two chapters to the volume on Global Catastrophic Risks). Heuristically, if I have a high opinion of Bostrom and the FHI project, that raises my opinion of Eliezer and decreases the probability of Eliezer-as-crackpot.
Well, in the category of "criticisms of SIAI and/or Eliezer", this text is certainly among the better ones. I could see this included on a "required reading list" of new SIAI employees or something.
But since we're talking about a Very Important Issue, i.e. existential risks, the text might have benefited from some closing warnings, that whatever people's perceptions of SIAI, it's Very Important that they don't neglect being very seriously interested in existential risks because of issues that they might perceive a particular organization working on the topic to have (and that it might also actually have, but that's not my focus in this comment).
I.e. if people think SIAI sucks and shouldn't be supported, they should anyway be very interested in supporting the Future of Humanity Institute at Oxford, for example. Otherwise they're demonstrating very high levels of irrationality, and with regard to SIAI, are probably just looking for plausible-sounding excuses to latch onto for why they shouldn't pitch in.
Not to say that the criticism you presented mightn't be very valid (or not; I'm not really commenting on that here), but it would be very important for people to f...
I don't think there's any point doing armchair diagnoses and accusing people of delusions of grandeur. I wouldn't go so far as to claim that Eliezer needs more self-doubt, in a psychological sense. That's an awfully personal statement to make publicly. It's not self-confidence I'm worried about, it's insularity.
Here's the thing. The whole SIAI project is not publicly affiliated with (as far as I've heard) other, more mainstream institutions with relevant expertise. Universities, government agencies, corporations. We don't have guest posts from Dr. X or Think Tank Fellow Y. The ideas related to friendly AI and existential risk have not been shopped to academia or evaluated by scientists in the usual way. So they're not being tested stringently enough.
It's speculative. It feels fuzzy to me -- I'm not an expert in AI, but I have some education in math, and things feel fuzzy around here.
If you want to claim you're working on a project that may save the world, fine. But there's got to be more to show for it, sooner or later, than speculative essays. At the very least, people worried about unfriendly AI will have to gather data and come up with some kind of statistical stu...
The whole SIAI project is not publicly affiliated with (as far as I've heard) other, more mainstream institutions with relevant expertise. Universities, government agencies, corporations. We don't have guest posts from Dr. X or Think Tank Fellow Y.
According to the about page, LW is brought to you by the Future of Humanity Institute at Oxford University. Does this count? Many Dr. Xes have spoken at the Singularity Summits.
At the very least, people worried about unfriendly AI will have to gather data and come up with some kind of statistical study that gives evidence of a threat!
It's not clear how one would use past data to give evidence for or against a UFAI threat in any straightforward way. There's various kinds of indirect evidence that could be presented, and SIAI has indeed been trying more in the last year or two to publish articles and give conference talks presenting such evidence.
Points that SIAI would do better if it had better PR, had more transparency, published more in the scientific literature, etc., are all well-taken, but these things use limited resources, which to me makes it sound strange to use them as arguments to direct funding elsewhere.
Here's the thing. The whole SIAI project is not publicly affiliated with (as far as I've heard) other, more mainstream institutions with relevant expertise.
LessWrong is itself a joint project of the SIAI and the Future of Humanity Institute at Oxford. Researchers at the SIAI have published these academic papers. The Singularity Summit's website includes a lengthy list of partners, including Google and Scientific American.
The SIAI and Eliezer may not have done the best possible job of engaging with the academic mainstream, but they haven't done a terrible one either, and accusations that they aren't trying are, so far as I am able to determine, factually inaccurate.
I feel that perhaps you haven't considered the best way to maximise your chance of developing Friendly AI if you were Eliezer Yudkowsky; your perspective is very much focussed on how you see it lookin in from the outside. Consider for a moment that you are in a situation where you think you can make a huge positive impact upon the world, and have founded an organisation to help you act upon that.
Your first, and biggest problem is getting paid. You could take time off to work on attaining a fortune through some other means but this is not a certain bet, and will waste years that you could be spending working on the problem instead. Your best bet is to find already wealthy people who can be convinced that you can change the world, that it's for the best, and that they should donate significant sums of money to you, unless you believe this is even less certain than making a fortune yourself. There's already a lot of people in the world with the requisite amount of money to spare. I think seeking donations is the more rational path.
Now, given that you need to persuade people of the importance of your brilliant new idea which no one has really been considering before, and that to most ...
How would you address this?
http://scienceblogs.com/pharyngula/2010/08/kurzweil_still_doesnt_understa.php
It seems to me like PZ Meyers really doesn't understand information theory. He's attacking Kurzweil and calling him a kook. Initially due to a relatively straightforward complexity estimate.
And I'm pretty confident that Myers is wrong on this, unless there is another information rich source of inheritance besides DNA, which Meyers knows about but Kurzweil and I do not.
This looks to me like a popular science blogger doing huge PR damage to everything singularity related, and being wrong about it. Even if he is later convinced of this point.
I don't see how to avoid this short of just holding back all claims which seem exceptional and that some 'reasonable' person might fail to understand and see as a sign of cultishness. If we can't make claims as basic as the design of the brain being in the genome, then we may as well just remain silent.
But then we wouldn't find out if we're wrong, and we're rationalists.
For instance, you can't measure the number of transistors in an Intel CPU and then announce, "A-ha! We now understand what a small amount of information is actually required to create all those operating systems and computer games and Microsoft Word, and it is much, much smaller than everyone is assuming."
This analogy made me cringe. Myers is disagreeing with the claim that human DNA completely encodes the structure and functioning of the human brain: the hardware and software, roughly. Looking at the complexity of the hardware and making claims about the complexity of the software, as he does here, is completely irrelevant to his disagreement. It serves only to obscure the actual point under debate, and demonstrates that he has no idea what he's talking about.
I upvoted this, but I'm torn about this.
In your recent posts you've been slowly, carefully, thoroughly deconstructing one person. Part of me wants to break into applause at the techniques used, and learn from them, because in my whole life of manipulation I've never mounted an attack of such scale. (The paragraph saying "something has gone very wrong" was absolutely epic, to the point of evoking musical cues somewhere at the edge of my hearing. Just like the "greatly misguided" bit in your previous post. Bravo!) But another part of me feels horror and disgust because after traumatic events in my own life I'd resolved to never do this thing again.
It comes down to this: I enjoy LW for now. If Eliezer insists on creating a sealed reality around himself, what's that to me? You don't have to slay every dragon you see. Saving one person from megalomania (real or imagined) is way less important than your own research. Imagine the worst possible world: Eliezer turns into a kook. What would that change, in the grand scheme of things or in your personal life? Are there not enough kooks in AI already?
And lastly, a note about saving people. I think many of us here have had ...
I saved someone from suicide once. While the experience was certainly quite unpleasant at the time, if I had hit "ignore," as you suggest, she would have died. I don't think that I would be better off today if I had let her die, to say nothing of her. The fact that saving people is hard doesn't mean that you shouldn't do it!
It comes down to this: I enjoy LW for now. If Eliezer insists on creating a sealed reality around himself, what's that to me? You don't have to slay every dragon you see. Saving one person from megalomania (real or imagined) is way less important than your own research. Imagine the worst possible world: Eliezer turns into a kook. What would that change, in the grand scheme of things or in your personal life?
The very fate of the universe, potentially. Purely hypothetically and for the sake of the discussion:
Er... I can't help but notice a certain humor in the idea that it's terrible if I'm self-deluded about my own importance because that means I might destroy the world.
It seems like an implication of your post that no one is ever allowed to believe they're saving the world. Do you agree that this is an implication? If not, why not?
Not speaking for multi, but, in any x-risk item (blowing up asteroids, stabilizing nuclear powers, global warming, catastrophic viral outbreak, climate change of whatever sort, FAI, whatever) for those working on the problem, there are degrees of realism:
"I am working on a project that may have massive effect on future society. While the chance that I specifically am a key person on the project are remote, given the fine minds at (Google/CDC/CIA/whatever), I still might be, and that's worth doing." - Probably sane, even if misguided.
"I am working on a project that may have massive effect on future society. I am the greatest mind in the field. Still, many other smart people are involved. The specific risk I am worried about may or not occur, but efforts to prevent its occurrence are valuable. There is some real possibility that I will the critical person on the project." - Possibly sane, even if misguided.
"I am working on a project that will save a near-infinite number of universes. In all likelihood, only I can achieve it. All of the people - even people perceived as having better credentials, intelligence, and ability - cannot do what I am doing. All crit...
I don't think Eliezer believes he's irreplaceable, exactly. He thinks, or I think he thinks, that any sufficiently intelligent AI which has not been built to the standard of Friendliness (as he defines it) is an existential risk. And the only practical means for preventing the development of UnFriendly AI is to develop superintelligent FAI first. The team to develop FAI needn't be SIAI, and Eliezer wouldn't necessarily be the most important contributor to the project, and SIAI might not ultimately be equal to the task. But if he's right about the risk and the solution, and his untimely demise were to doom the world, it would be because no-one else tried to do this, not because he was the only one who could.
Not that this rules out your interpretation. I'm sure he has a high opinion of his abilities as well. Any accusation of hubris should probably mention that he once told Aubrey de Grey "I bet I can solve ALL of Earth's emergency problems before you cure aging."
I certainly do not assign a probability as high as 70% to the conjunction of all three of those statements.
And in case it wasn't clear, the problem I was trying to point out was simply with having forbidden conclusions - not forbidden by observation per se, but forbidden by forbidden psychology - and using that to make deductions about empirical premises that ought simply to be evaluated by themselves.
I s'pose I might be crazy, but you all are putting your craziness right up front. You can't extract milk from a stone!
I assign a probability of less than 10^(-9) to you succeeding in playing a critical role on the Friendly AI project that you're working on.
I wish the laws of argument permitted me to declare that you had blown yourself up at this point, and that I could take my toys and go home. Alas, arguments are not won on a points system.
My impression is that you've greatly underestimated the difficulty of building a Friendly AI.
Out of weary curiosity, what is it that you think you know about Friendly AI that I don't?
And has it occurred to you that if I have different non-crazy beliefs about Friendly AI then my final conclusions might not be so crazy either, no matter what patterns they match in your craziness recognition systems?
I agree it's kind of ironic that multi has such an overconfident probability assignment right after criticizing you for being overconfident. I was quite disappointed with his response here.
One could offer many crude back-of-envelope probability calculations. Here's one: let's say there's
That seems conservative to me and implies at least a 1 in 10^4 chance. Obviously there's lots of room for quibbling here, but it's hard for me to see how such quibbling could account for five orders of magnitude. And even if post-quibbling you think you have a better model that does imply 1 in 10^9, you only need to put little probability mass on my model or models like it for them to dominate the calculation. (E.g., a 9 in 10 chance of a 1 in 10^9 chance plus a 1 in 10 chance of a 1 in 10^4 chance is close to a 1 in 10^5 chance.)
Let's pick an example. How probable do you think it is that Islam is a true religion? (There are several ways to take care of logical contradictions here, so saying 0% is not an option.)
Suppose there were a machine--for the sake of tradition, we can call it Omega--that prints out a series of zeros and ones according to the following rule. If Islam is true, it prints out a 1 on each round, with 100% probability. If Islam is false, it prints out a 0 or a 1, each with 50% probability.
Let's run the machine... suppose on the first round, it prints out a 1. Then another. Then another. Then another... and so on... it's printed out 10 1's now. Of course, this isn't so improbable. After all, there was a 1/1024 chance of it doing this anyway, even if Islam is false. And presumably we think Islam is more likely than this to be false, so there's a good chance we'll see a 0 in the next round or two...
But it prints out another 1. Then another. Then another... and so on... It's printed out 20 of them. Incredible! But we're still holding out. After all, million to one chances happen every day...
Then it prints out another, and another... it just keeps going... It's printed out 30 1's now. Of course...
Thank you a lot for posting this scenario. It's instructive from the "heuristics and biases" point of view.
Imagine there are a trillion variants of Islam, differing by one paragraph in the holy book or something. At most one of them can be true. You pick one variant at random, test it with your machine and get 30 1's in a row. Now you should be damn convinced that you picked the true one, right? Wrong. Getting this result by a fluke is 1000x more likely than having picked the true variant in the first place. Probability is unintuitive and our brains are mush, that's all I'm sayin'.
I wish the laws of argument permitted me to declare that you had blown yourself up at this point, and that I could take my toys and go home. Alas, arguments are not won on a points system.
On the other hand, assuming he knows what it means to assign something a 10^-9 probability, it sounds like he's offering you a bet at 1000000000:1 odds in your favour. It's a good deal, you should take it.
So, my impression is that you and Eliezer have different views of this matter. My impression is that Eliezer's goal is for SIAI to actually build an AGI unilaterally.
Still, build AGI eventually, and not now. Expertise in AI/AGI is of low relevance at present.
It seems much more feasible to develop a definition of friendliness and then get governments to mandate that it be implemented in any AI or something like that.
It seems obviously infeasible to me that governments will chance upon this level of rationality. Also, we are clearly not on the same page if you say things like "implement in any AI". Friendliness is not to be "installed in AIs", Friendliness is the AI (modulo initial optimizations necessary to get the algorithm going and self-optimizing, however fast or slow that's possible). The AGI part of FAI is exclusively about optimizing the definition of Friendliness (as an algorithm), not about building individual AIs with standardized goals.
See also this post for a longer explanation of why weak-minded AIs are not fit to carry the definition of Friendliness. In short, such AIs are (in principle) as much an existential danger as human AI researchers.
The 1-in-a-billion follows not from it being plausible that there are three such events, but from it being virtually certain. Models without such events will end up dominating the final probability. I can easily imagine that if I magically happened upon a very reliable understanding of some factors relevant to future FAI development, the 1 in a billion figure would be the right thing to believe. But I can easily imagine it going the other way, and absent such understanding, I have to use estimates much less extreme than that.
FWIW, as an entrepreneur type I consider one of my top 3 key advantages the fact that I would actually appreciate it greatly if someone explained in detail why I was wasting my time with my current project. Thinking about this motivates me significantly because I haven't met any other entrepreneur types who I'd guess this is also true for.
The real bone of contention here seems to be the long chain of inference leading from common scientific/philosophical knowledge to the conclusion that uFAI is a serious existential risk. Any particular personal characteristics of EY would seem irrelevant till we have an opinion on that set of claims.
If EY were working on preventing asteroid impacts with earth, and he were the main driving force behind that effort, he could say "I'm trying to save the world" and nobody would look at him askance. That's because asteroid impacts have definitely caused mass extinctions before, so nobody can challenge the very root of his claim.
The FAI problem, on the other hand, is at the top of a large house of inferential cards, so that Eliezer is saving the world GIVEN that W, X, Y and Z are true.
My bottom line: what we should be discussing is simply "Are W, X, Y and Z true?" Once we have a good idea about how strong that house of cards is, it will be obvious whether Eliezer is in a "permissible" epistemic state, or whatever.
Maybe people who know about these questions should consider a series of posts detailing all the separate issues leading to FAI. As far as I can tell from my not-extremely-tech-savvy vantage point, the weakest pillar in that house is the question of whether strong AI is feasible (note I said "feasible," not "possible").
An interesting post, well written, upvoted. Mere existence of such posts here constitutes a proof that LW is still far from Objectivism, not only because Eliezer is way more rational (and compassionate) than Ayn Rand, but mainly because the other people here are aware of dangers of cultism.
However, I am not sure whether the right way to prevent cultish behaviour (whether the risk is real or not) is to issue warning like this to the leader (or any sort of warning, perhaps). The dangers of cultism emerge from simply having a leader; whatever the level of personal rationality, being a single extraordinarily revered person in any group for any longer time probably harm's one's judgement, and the overall atmosphere of reverence is unhealthy for the group. Maybe more generally, the problem not necessarily depends on existence of a leader: if a group is too devoted to some single idea, it faces lots of dangers, the gravest thereof perhaps be separation from reality. Especially if the idea lives in an environment where relevant information is not abundant.
Therefore, I would prefer to see the community concentrate on a broader class of topics, and to continue in the tradition of disseminati...
I find it ironic that multifoliaterose said
I personally think that the best way to face the present situation is to gather more information about all existential risks rather than focusing on one particular existential risk
and then the next post, instead of delineating what he found out about other existential risks (or perhaps how we should go about doing that), is about how to save Eliezer.
As of yet Eliezer's importance is just a stochastic variable yet to be realized, for all I know he could be killed in a car accident tomorrow or simply fail at his task of "saving the world" in numerous ways.
Up until now Vasili Arkhipov, Stanislav Petrov and a few other people I do not know the names of (including our earliest ancestors who managed to avoid being killed during their emigration out of Africa) trump Eliezer by a tiny margin of actually saving humanity -or at least civilization.
All that being said Eliezer is still pretty awesome by my standards. And he writes good fanfiction, too.
Who else is nearly as good or better at Friendly AI development than Eliezer Yudkowsky?
I mean besides me, obviously.
it's very important that those of us who aspire to epistemic rationality incorporate a significant element of "I'm the sort of person who engages in self-doubt because it's the right thing to do" into our self-image
I think most of us do. Your argument for this is compelling. However, I think Eliezer was just claiming that it's possible to overdo it - at least, that's the defensible core of his insight.
I've wondered if I'm obsessed with Eliezer's writings, and whether I esteem him too highly. Answers: no, and no.
...Anything that has even a sl
...Take no pride in your confession that you too are biased; do not glory in your self-awareness of your flaws. This is akin to the principle of not taking pride in confessing your ignorance; for if your ignorance is a source of pride to you, you may become loathe to relinquish your ignorance when evidence comes knocking. Likewise with our flaws - we should not gloat over how self-aware we are for confessing them; the occasion for rejoicing is when we have a little less to confess.
There's something to what Eliezer is saying here: when people are too str
This post is a pretty accurate description of me a few years ago, when I was a Singularitarian. The largest attraction of the belief system, to me, was that it implied as an AI researcher I was not just a hero, but a superhero, potentially capable of almost single-handedly saving the world. (And yes, I loved those video games too.)
I have no comment to add but I will say that this is well written and researched. It also prompted a degree of self reflection on my part. At least, that's what I told myself and I feel this warm glow inside. ;)
As far as I can tell, Eliezer does have confidence in the idea that he is (at least nearly) the most important person in human history. Eliezer's silence only serves to further confirm my earlier impressions
I suppose you also believe that Obama must prove he's not a muslim? And must do so again every time someone asserts that he is?
Let me say that Eliezer may have already done more to save the world than most people in history. This is going on the assumption that FAI is a serious existential risk. Even if he is doing it wrong and his work will never di...
Honestly, I don't think Eliezer would look overly eccentric if it weren't for LessWrong/Overcomingbias. Comp sci is notoriously eccentric, AI research possibly more so. The stigma against Eliezer isn't from his ideas, it isn't from his self confidence, it's from his following.
Kurzweil is a more dulled case: he has good ideas, but is clearly sensational, he has a large following, but that following isn't nearly as dedicated as the one to Eliezer (not necessarily to Eliezer himself, but to his writings and the "practicing of rationality"). And the ...
A few unrelated points:
I first came across Overcoming Bias in 2008. Eliezer was recommended to me by a friend who I respect a great deal. My reactions to the first postings that I read by Eliezer was strong discomfort with his apparent grandiosity and self absorption. This discomfort was sufficiently strong for me to lose interest despite my friend's endorsement.
I'd be really interested to know which posts these were, because it would help me to distinguish between the following interpretations:
(1) First impressions really do matter: even though you and I are probably very similar in many respects, we have different opinions of Eliezer simply because in the first posts of his I read, he sounded more like a yoga instructor than a cult leader; whereas perhaps the first thing you read was some post where his high estimation of his abilities relative to the rest of humanity was made explicit, and you didn't have the experience of his other writings to allow you to "forgive" him for this social transgression.
(2) We have different personalities, which cause us to interpret people's words differently: you and I read more or less the same kind of material first, but you just interpreted it as "...
In high school I went through a period when I believed that I was a messianic figure whose existence had been preordained by a watchmaker God who planned for me to save the human race. It's appropriate to say that during this period of time I suffered from extreme delusions of grandeur. I viscerally understand how it's possible to fall into an affective death spiral.
Not that the two are exclusive, but this sounds an awful lot like a manic episode. I assume you gave that due consideration?
The mechanism that determines human action is that we do what makes us feel good (at the margin) and refrain from doing what makes us feel bad (at the margin).
"The" mechanism? Citation needed.
a fundamental mechanism of the human brain which was historically correlated with gaining high status is to make us feel good when we have high self-image and feel bad when we have low self-image.
Better, but still unsupported and unclear. What was correlated with what?
I'm inclined to think that Eliezer's confidence in his own importance (actually I'd prefer "expected importance", measured as an expected quantity the usual way) is not really unwarranted (or too unwarranted), but I hope he does take away from this a greater sense of the importance of a "the customer is always right" attitude in managing his image as a public-ish figure. Obviously the customer is not always right, but sometimes you have to act like they are if you want to get/keep them as your customer... justified or not, there seems t...
we do what makes us feel good (at the margin) and refrain from doing what makes us feel bad (at the margin)
By "(at the margin)" do you just mean that we're definitely aiming at some local optimum, and possibly missing better (but harder to reach) world/self states? If so, I agree. If not, I don't get your meaning.
[Added 02/24/14: After I got feedback on this post, I realized that it carried unnecessary negative connotations (despite conscious effort on my part to avoid them), and if I were to write it again, I would have framed things differently. See Reflections on a Personal Public Relations Failure: A Lesson in Communication for more information. SIAI (now MIRI) has evolved substantially since 2010 when I wrote this post, and the criticisms made in the post don't apply to MIRI as presently constituted.]
Follow-up to: Other Existential Risks, Existential Risk and Public Relations
Related to: Tsuyoku Naritai! (I Want To Become Stronger), Affective Death Spirals, The Proper Use of Doubt, Resist the Happy Death Spiral, The Sin of Underconfidence
In Other Existential Risks I began my critical analysis of what I understand to be SIAI's most basic claims. In particular I evaluated part of the claim
(1) At the margin, the best way for an organization with SIAI's resources to prevent global existential catastrophe is to promote research on friendly Artificial Intelligence, work against unsafe Artificial Intelligence, and encourage rational thought.
It's become clear to me that before I evaluate the claim
(2) Donating to SIAI is the most cost-effective way for charitable donors to reduce existential risk.
I should (a) articulate my reasons for believing in the importance of self-doubt and (b) give the SIAI staff an opportunity to respond to the points which I raise in the present post as well as my two posts titled Existential Risk and Public Relations and Other Existential Risks.
Yesterday SarahC described to me how she had found Eliezer's post Tsuyoku Naritai! (I Want To Become Stronger) really moving. She explained:
I thought it was good: the notion that you can and must improve yourself, and that you can get farther than you think.
I'm used to the other direction: "humility is the best virtue."
I mean, this is a big fuck-you to the book of Job, and it appeals to me.
I was happy to learn that SarahC had been positively affected by Eliezer's post. Self-actualization is a wonderful thing and it appears as though Eliezer's posting has helped her self-actualize. On the other hand, rereading the post prompted me to notice that there's something about it which I find very problematic. The last few paragraphs of the post read:
Take no pride in your confession that you too are biased; do not glory in your self-awareness of your flaws. This is akin to the principle of not taking pride in confessing your ignorance; for if your ignorance is a source of pride to you, you may become loathe to relinquish your ignorance when evidence comes knocking. Likewise with our flaws - we should not gloat over how self-aware we are for confessing them; the occasion for rejoicing is when we have a little less to confess.
Otherwise, when the one comes to us with a plan for correcting the bias, we will snarl, "Do you think to set yourself above us?" We will shake our heads sadly and say, "You must not be very self-aware."
Never confess to me that you are just as flawed as I am unless you can tell me what you plan to do about it. Afterward you will still have plenty of flaws left, but that's not the point; the important thing is to do better, to keep moving ahead, to take one more step forward. Tsuyoku naritai!
There's something to what Eliezer is saying here: when people are too strongly committed to the idea that humans are fallible this can become a self-fulfilling prophecy where humans give up on trying to improve things and as a consequence remain fallible when they could have improved. As Eliezer has said in The Sin of Underconfidence, there are social pressures that push against having high levels of confidence even when confidence is epistemically justified:
To place yourself too high - to overreach your proper place - to think too much of yourself - to put yourself forward - to put down your fellows by implicit comparison - and the consequences of humiliation and being cast down, perhaps publicly - are these not loathesome and fearsome things?
To be too modest - seems lighter by comparison; it wouldn't be so humiliating to be called on it publicly, indeed, finding out that you're better than you imagined might come as a warm surprise; and to put yourself down, and others implicitly above, has a positive tinge of niceness about it, it's the sort of thing that Gandalf would do.
I have personal experience with underconfidence. I'm a careful thinker and when I express a position with confidence my position is typically well considered. For many years I generalized from one example and assumed when people express positions with confidence they've thought their positions out as well as I have. Even after being presented with massive evidence that few people think things through as carefully as I do, I persisted in granting the (statistically ill-considered) positions of others far more weight than they deserved for the very reason that Eliezer describes above. This seriously distorted my epistemology because it led to me systematically giving ill-considered positions substantial weight. I feel that I have improved on this point, but even now, from time to time I notice that I'm exhibiting irrationally low levels of confidence in my positions.
At the same time, I know that at times I've been overconfident as well. In high school I went through a period when I believed that I was a messianic figure whose existence had been preordained by a watchmaker God who planned for me to save the human race. It's appropriate to say that during this period of time I suffered from extreme delusions of grandeur. I viscerally understand how it's possible to fall into an affective death spiral.
In my view one of the central challenges of being human is to find an instrumentally rational balance between subjecting oneself to influences which push one in the direction of overconfidence and subjecting oneself to influences which push one in the direction of underconfidence.
In Tsuyoku Naritai! Eliezer describes how Orthodox Judaism attaches an unhealthy moral significance to humility. Having grown up in a Jewish household and as a consequence having had peripheral acquaintance with orthodox Judaism I agree with Eliezer's analysis of Orthodox Judaism in this regard. In the proper use of doubt, Eliezer describes how the Jesuits allegedly are told to doubt their doubts about Catholicism. I agree with Eliezer that self-doubt can be misguided and abused.
However, reversed stupidity is not intelligence. The fact that it's possible to ascribe too much moral significance to self-doubt and humility does not mean that one should not attach moral significance to self-doubt and humility. I strongly disagree with Eliezer's prescription: "Take no pride in your confession that you too are biased; do not glory in your self-awareness of your flaws."
The mechanism that determines human action is that we do what makes us feel good (at the margin) and refrain from doing what makes us feel bad (at the margin). This principle applies to all humans, from Gandhi to Hilter. Our ethical challenge is to shape what makes us feel good and what makes us feel bad in a way that incentivizes us to behave in accordance with our values. There are times when it's important to recognize that we're biased and flawed. Under such circumstances, we should feel proud that we recognize that we're biased we should glory in our self-awareness of our flaws. If we don't, then we will have no incentive to recognize that we're biased and be aware of our flaws.
We did not evolve to exhibit admirable and noble behavior. We evolved to exhibit behaviors which have historically been correlated with maximizing our reproductive success. Because our ancestral climate was very much a zero-sum situation, the traits that were historically correlated with maximizing our reproductive success had a lot to do with gaining high status within our communities. As Yvain has said, it appears that a fundamental mechanism of the human brain which was historically correlated with gaining high status is to make us feel good when we have high self-image and feel bad when we have low self-image.
When we obtain new data, we fit it into a narrative which makes us feel as good about ourselves as possible; a way conducive to having a high self-image. This mode of cognition can lead to very seriously distorted epistemology. This is what happened to me in high school when I believed that I was a messianic figure sent by a watchmaker God. Because we flatter ourselves by default, it's very important that those of us who aspire to epistemic rationality incorporate a significant element of "I'm the sort of person who engages in self-doubt because it's the right thing to do" into our self-image. If we do this, when we're presented with evidence which entails a drop in our self-esteem, we don't reject it out of hand or minimize it as we've been evolutionarily conditioned to do because wound of properly assimilating data is counterbalanced by the salve of the feeling "At least I'm a good person as evidenced by the fact that I engage in self-doubt" and failing to exhibit self-doubt would itself entail an emotional wound.
This is the only potential immunization to the disease of self-serving narratives which afflicts all utilitarians out of virtue of their being human. Until technology allows us to modify ourselves in a radical way, we cannot hope to be rational without attaching moral significance to the practice of engaging in self-doubt. As the RationalWiki's page on LessWrong says:
A common way for very smart people to be stupid is to think they can think their way out of being apes with pretensions. However, there is no hack that transcends being human...You are an ape with pretensions. Playing a "let's pretend" game otherwise doesn't mean you win all arguments, or any. Even if it's a very elaborate one, you won't transcend being an ape. Any "rationalism" that doesn't expressly take into account humans being apes with pretensions, isn't.
In Existential Risk and Public Relations I suggested that some of Eliezer's remarks convey the impression that Eliezer has an unjustifiably high opinion of himself. In the comments to the post JRMayne wrote
I think the statements that indicate that [Eliezer] is the most important person in human history - and that seems to me to be what he's saying - are so seriously mistaken, and made with such a high confidence level, as to massively reduce my estimated likelihood that SIAI is going to be productive at all.
And that's a good thing. Throwing money into a seriously suboptimal project is a bad idea. SIAI may be good at getting out the word of existential risk (and I do think existential risk is serious, under-discussed business), but the indicators are that it's not going to solve it. I won't give to SIAI if Eliezer stops saying these things, because it appears he'll still be thinking those things.
When Eliezer responded to JRMayne's comment, Eliezer did not dispute the claim that JRMayne attributed to him. I responded to Eliezer saying
If JRMayne has misunderstood you, you can effectively deal with the situation by making a public statement about what you meant to convey.
Note that you have not made a disclaimer which rules out the possibility that you claim that you're the most important person in human history. I encourage you to make such a disclaimer if JRMayne has misunderstood you.
I was disappointed, but not surprised, that Eliezer did not respond. As far as I can tell, Eliezer does have confidence in the idea that he is (at least nearly) the most important person in human history. Eliezer's silence only serves to further confirm my earlier impressions. I hope that Eliezer subsequently proves me wrong. [Edit: As Airedale points out Eliezer has in fact exhibited public self-doubt in his abilities in his posting The Level Above Mine. I find this reassuring and it significantly lowers my confidence that Eliezer claims that he's the most important person in human history. But Eliezer still hasn't made a disclaimer on this matter decisively indicating that he does not hold such a view.] The modern world is sufficiently complicated so that no human no matter how talented can have good reason to believe himself or herself to be the most important person in human history without actually doing something which very visibly and decisively alters the fate of humanity. At present, anybody who holds such a belief is suffering from extreme delusions of grandeur.
There's some sort of serious problem with the present situation. I don't know whether it's a public relations problem or if the situation is that Eliezer actually suffers from extreme delusions of grandeur, but something has gone very wrong. The majority of the people who I know who outside of Less Wrong who have heard of Eliezer and Less Wrong have the impression that Eliezer is suffering from extreme delusions of grandeur. To such people, this fact (quite reasonably) calls into question of the value of SIAI and Less Wrong. On one hand, SIAI looks like an organization which is operating under beliefs which Eliezer has constructed to place himself in as favorable a position as possible rather than with a view toward reducing existential risk. On the other hand, Less Wrong looks suspiciously like the cult of Objectivism: a group of smart people who are obsessed with the writings of a very smart person who is severely deluded and describing these writings and the associated ideology as "rational" although they are nothing of the kind.
My own views are somewhat more moderate. I think that the Less Wrong community and Eliezer are considerably more rational than the Objectivist movement and Ayn Rand (respectively). I nevertheless perceive unsettling parallels.
In the comments to Existential Risk and Public Relations, timtyler said
...many people have inflated views of their own importance. Humans are built that way. For one thing, It helps them get hired, if they claim that they can do the job. It is sometimes funny - but surely not a big deal.
I disagree with timtyler. Anything that has even a slight systematic negative impact on existential risk is a big deal.
Some of my most enjoyable childhood experiences involved playing Squaresoft RPGs. Games like Chrono Trigger, Illusion of Gaia, Earthbound, Xenogears, and the Final Fantasy series are all stories about a group of characters who bond and work together to save the world. I found these games very moving and inspiring. They prompted me to fantasize about meeting allies who I could bond with and work together with to save the world. I was lucky enough to meet one such person in high school who I've been friends with since. When I first encountered Eliezer I found him eerily familiar, as though he was a long lost brother. This is the same feeling that is present between Siegmund and Sieglinde in the Act 1 of Wagner's Die Walküre (modulo erotic connotations). I wish that I could be with Eliezer in a group of characters as in a Squaresoft RPG working to save the world. His writings such as One Life Against the World and Yehuda Yudkowsky, 1985-2004 reveal him to be a deeply humane and compassionate person.
This is why it's so painful for me to observe that Eliezer appears to be deviating so sharply from leading a genuinely utilitarian lifestyle. I feel a sense of mono no aware, wondering how things could have been under different circumstances.
One of my favorite authors is Kazuo Ishiguro, who writes about the themes of self-deception and people's attempts to contribute to society. In a very good interview Ishiguro said
I think that's partly what interests me in people, that we don't just wish to feed and sleep and reproduce then die like cows or sheep. Even if they're gangsters, they seem to want to tell themselves they're good gangsters and they're loyal gangsters, they've fulfilled their 'gangstership' well. We do seem to have this moral sense, however it's applied, whatever we think. We don't seem satisfied, unless we can tell ourselves by some criteria that we have done it well and we haven't wasted it and we've contributed well. So that is one of the things, I think, that distinguishes human beings, as far as I can see.
But so often I've been tracking that instinct we have and actually looking at how difficult it is to fulfill that agenda, because at the same time as being equipped with this kind of instinct, we're not actually equipped. Most of us are not equipped with any vast insight into the world around us. We have a tendency to go with the herd and not be able to see beyond our little patch, and so it is often our fate that we're at the mercy of larger forces that we can't understand. We just do our little thing and hope it works out. So I think a lot of the themes of obligation and so on come from that. This instinct seems to me a kind of a basic thing that's interesting about human beings. The sad thing is that sometimes human beings think they're like that, and they get self-righteous about it, but often, they're not actually contributing to anything they would approve of anyway.
[...]
There is something poignant in that realization: recognizing that an individual's life is very short, and if you mess it up once, that's probably it. But nevertheless, being able to at least take some comfort from the fact that the next generation will benefit from those mistakes. It's that kind of poignancy, that sort of balance between feeling defeated but nevertheless trying to find reason to feel some kind of qualified optimism. That's always the note I like to end on. There are some ways that, as the writer, I think there is something sadly pathetic but also quite noble about this human capacity to dredge up some hope when really it's all over. I mean, it's amazing how people find courage in the most defeated situations.
Ishiguro's quote describes how people often behave in accordance with sincere desire to contribute and end up doing things that are very different from what they thought they were doing (things which are relatively unproductive or even counterproductive). Like Ishiguro I find this phenomenon very sad. As Ishiguro hints at, this phenomenon can also result in crushing disappointment later in life. I feel a deep spiritual desire to prevent this from happening to Eliezer.
You may want to read this paper I presented at FHI. Note that there's a big difference between the probability of risk conditional on WBE coming first or AI coming first and marginal impact of effort. In particular some of our uncertainty is about logical facts about the space of algorithms and technology landscape, and some of it is about the extent and effectiveness of activism/intervention.
Thanks for the very interesting reference! Is it linked on the SIAI research papers page? I didn't see it there.
Note that there's a big difference between the probability of risk conditional on WBE coming first or AI coming first and marginal impact of effort.
I appreciate this point which you've made to me previously (and which appears in your comment that I linked above!).