If you solve the problem of human-friendly self-improving AI, you have indirectly solved every problem. After spending a decent amount of time on LW, I have been convinced of this premise and now I would like to devote my life to that cause.

 

Currently I'm living in Germany and I'm studying psychology in the first semester. The university I'm studying at has a great reputation (even internationally if I can believe the rankings) for the quality of its scientific psychology research and it ranks about #2nd or #3rd place when it comes to various psy-science-related criteria out of about 55 German universities where one can study psychology. Five semesters of statistics in my Bachelor of Science might also hint at that.

I want to finish my Bachelor of Science and then move on to my Master, so in about 5 years I might hit my "phase of actual productivity" in the working field. I'm flirting with cognitive neuroscience, but haven't made my decision yet - however, I am pretty sure that I want to move towards research and a scientific career rather than one in a therapeutic field.

Before discovering lesswrong my most dominant personal interest in psychology has been in the field of "positive psychology" or plainly speaking the "what makes humans happy" field. This interest hasn't really changed through the discovery of LW, as much as it has evolved into: "how can we distill what makes human life worthwhile and put it into terms a machine could execute for our benefit"?

 

As the title suggests, I'm writing all this because I want some creative input from you in order to expand my sense of possibilities concerning how I can help the development of friendly AI from the field of psychology most effectively.

 

To give you a better idea of what might fit me, a bit more background-info about myself and my abilities seems in order:

I like talking and writing a lot, mathematically I am a loser (whether due to early disgust or incompetence I can't really tell). I value and enjoy human contact and have constantly moved from being an introvert towards being an extrovert by several cognitive developments I can only speculate on. I would probably easily rank in the middle field of any positive extroversion scale nowadays. My IQ seems to be around 134 if one can trust the "International High IQ Society" (www.highiqsociety.org), but as mentioned my abilities probably lie more in the linguistic and to some extent analytic sphere than the mathematical. I understand Bayes' Theorem but haven't read the quantum mechanics sequence and many "higher" concepts here are still above my current level of comprehension. Although I haven't tried all that hard yet to be fair.

I have programmed some primitive HTML and CSS once and didn't really like it. From that experiecne and my mathematical inability I take away, that programming wouldn't be the way that I could contribute most efficiently towards friendly AI-research. It is none of my strenghts or at least it would take a lot of time to develop that, which would probably be better spent somewhere else. Also I quite surely wouldn't enjoy it as much as work in the psychological realm with humans.

My English is almost indistinguishable from that of a native speaker and I largely lack that (rightfully) despised and annoying German accent, so I could definitely see myself giving competent talks in English.

Like many of you I have serious problems with akrasia (regardless of whether that's a rationalist phenomenon or whether we are just more aware of it and tend to do types of work that tempt it more readily). Before I learned of how to effectively combat it (thank you Piers Steel!), I had plenty of motivation to get rid of it and sunk insane efforts into overcoming it, although ultimately it was largely an unsuccessful undertaking due to half-assed pop-science and the lack of a real insight about what procrastination is caused by and how it actually functions. Now that I know how to fix procrastination (or rather now that I know that it can't be fixed, as much as it has to be managed in a similar fashion to any given drug-addition), my motivation to overcome it is almost gone and I feel myself slacking. Also, the high certainty that there is no such thing as "free will" may have played a serious part in my procrastination habits (interestingly, there are at least two papers I recall showing this correlation). In a nutshell: Procrastination is a problem that I need to address, since it is definitely the Achilles' heel of my performance and it's absolutely crippling my potential. I probably rank middle-high on the impulsiveness- (and thus also on the procrastination-) scale.

That should be an adequate characterization of myself for now.

 

I am absolutely open for suggestions that are not related to the neuroscience of "what makes humans happy and how do I distill those goals and feelings into something a machine could work with"-field, but currently I am definitely flirting with that idea, even though I have absolutely no clue how the heck this area of research could be sufficiently financed in a decade from now and how it could spit out findings precise enough to benefit the creation of FAI. Yet maybe it's just a lack of imagination.

Trying to help set up and evolve a rationalist community in Germany would also be a decent task, but compared to specific research that actually directly aids our goals... I somehow feel it is less than what I could reasonably achieve if I really set my mind to it.

 

So tell me, where does a German psychologist go nowadays to achieve the biggest possible positive impact in the field of friendly AI?

New to LessWrong?

New Comment
42 comments, sorted by Click to highlight new comments since: Today at 1:43 PM

By priority:

  1. Do whatever makes the most money so you can donate to existential risk reduction.
  2. Study neuroeconomics (potentially useful for the 'friendliness content' problem).
  3. Study debiasing methods (in particular, build on the work of people like Keith Stanovich and Richard Larrick ).

You are in exactly the same position as me. Everything in your post reflects facts about my life, except a few:

  • I am not an extrovert, even in the slightest
  • I live in Florida, which isn't in Germany at all.
  • I've never written any code, even basic HTML

I don't have a fucking clue what I should be doing. It looks right now that I'm not going to make any difference at all. Not going to make a lot of money, definitely not going to contribute directly to FAI research. My current plan (which sucks) is to finish reading the Sequences and then apply for the existential risk career network. If I fail to get in, my Plan B is to come up with another plan.

I'm so helpless!

The existential risk career network is a loose mailing list. I would be surprised if you weren't accepted. But joining it won't solve your problems either...

I live in Florida, which isn't in Germany at all.

Huh. Well, ain't that somethin'.

I currently don't believe that psychology/cognitive science/neuroscience can be directly helpful for advancing the state of FAI, it seems more like a philosophy/math problem. It could be strategically helpful in developing arguments for complexity of human value. Other than that, if psychology is a necessary constraint, there's money, the unit of caring, so look for higher-income directions.

I'm thinking of majoring in Math and minoring in computational neuroscience if I get into my top choice university. Is it your opinion that that minor would be a waste of time and effort?

If you expect to be able to directly contribute to FAI, my current very tentative advice (which I'm following myself) is to focus on pure math, assuming you understand the philosophical side (problem statement) well enough (which would be the limiting factor for people who are already mathematicians). Take this with a lot of salt, since it's more of a hunch following from the character of seemingly relevant informal ideas, and I can't exhibit actual FAI-relevant math (other than the decision theory that was previously published, which is an important foothold and whose meaning I still don't quite grasp, which is also one motivation).

I currently don't believe that psychology/cognitive science/neuroscience can be directly helpful for advancing the state of FAI

I'm much less convinced about this. Knowing more about how the human brain actually measures value and what value is for humans seems hugely useful, and some of the stuff in that direction that lukeprog has been posting has seemed potentially very helpful. It does not seem implausible to presume that FAI is simply impossible without that knowledge.

Yeah. This is one of the issues on which my intuitions seem to differ the most with Nesov's, but I think we each concede that the other might be correct.

Right. If the philosophy/math path turns out to be genuinely impossible in some sense, neuroscience, if correctly interpreted, would be the next best source of detailed knowledge about human value. (I don't expect this to be feasible without a human-powered WBE singleton vastly stretching the available time before a global catastrophe.)

I'm confused. Isn't "detailed knowledge about human value" the part of the problem that we'd hand off to an AI implementing CEV?

What do you mean? Are you objecting to seeing a non-math path as even a barely feasible direction (i.e. it must be something handable to an AI, hence not neuroscience), or to the necessity of solving this problem when we already have CEV (i.e. all we need is to hand this problem over to an AI)? Both interpretations seem surprising, so I probably didn't catch the right one.

Closer to the latter, but I'm not really objecting to anything, so much as just being confused. I assume you're not saying we have to map out the "thousand shards of desire" before running CEV. I'm not saying there aren't various aspects of human value that would be useful to understand before we run CEV, but I'm curious what sort of details you had in mind, such that you think they could be figured out with neuroscience and without philosophy but such that you expect doing so to take a very long time.

CEV is way too far from a useful design, or even a big-picture sketch. It's a non-technical vague description of something that doesn't automatically fail for obvious reasons, as compared to essentially all other descriptions of human decision problem (friendliness content) published elsewhere. But "running CEV" is like running the sketch of da Vinci's flying machine.

CEV is a reasonable description on the level where a sketch of a plane doesn't insist on the plane having a beak or being made entirely out of feathers. It does look very good in comparison to other published sketches. But we don't have laws of aerodynamics or wind tunnels. "Running the sketch" is never a plan (which provokes protestations such as this). One (much preferable) way of going forward is to figure out the fundamental laws, that's decision theory/philosophy/math path. Another is to copy a bird in some sense, collecting all of its properties in as much detail as possible (metaphorically speaking, so that it's about copying goals and not about emulating brains); that's the neuroscience path, which I expect isn't viable no matter how much time is given, since we don't really know how to learn about goals by looking at brains or behavior.

(Perhaps when we figure out the fundamental laws, it'll turn out that we want a helicopter, and to the dustbin goes the original sketch.)

I agree that CEV needs conceptual and technical fleshing-out; when I said "run CEV", I meant "run some suitably fleshed-out version of CEV". You seem to be saying that to do this fleshing-out, we will need knowledge of some large subset of the details of human value. I'm not saying that's false, but I'm trying to get at what sort of details you think those are; what variables we're trying to find out the value of. Again, surely it's not all the details, or we wouldn't need to run CEV in the first place.

I do hope the value problem turns out to be solvable with philosophy/math instead of cognitive science. The philosophy/math path is much preferred.

Sorry if this has been discussed before, but isn't an open question whether WBE or AGI comes first? Couldn't WBE be a path to AGI?

I'm fairly sure WBE will have rather limited usefulness for AGI. The human brain is an evolved organ that looks like a patchwork and it isn't streamlined for intelligence - our numerous biases give testimony to that. There are a million mistakes we're making while we think (or rather while we think we think) and we're crap at things like self-control and giving reason preference over our feelings.

The human brain is deeply obsessed with social status and it's a hypocrite with antisocial tendencies when it knows that no one is looking. Emulating a whole human brain and somehow adding more speed or some kind of über-rationality module doesn't sound like a clean design/promising base for AGI at all. I for one wouldn't trust it.

Pursuing WBE for the purpose of building AGI is like trying to replicate a bird's feathers and movements in order to get people into the air, instead of building an airplane which is better suited for that purpose. Artificial intelligence probably won't need to borrow all that much from the way biological brains work.

EDIT: Actually, I just realized that you wrote AGI, not FAI (friendly AI). If your goal is to emulate human-level intelligence along with emotions and everything else a human brain entails, then WBE is the way to go for AGI (General Intelligence). You could also play around and adapt the emulation to suit some specific purpose, but that has serious ethical and safety-ramifications. If you are talking about creating a maximally intelligent agent, then you can probably do better than building on top of WBE.

From an utilitarian point of view, why would you want to risk creating an uFAI if you have WBE and hence virtual reality and hence the ability to create a utopia? Is it worth the risks?

Sorry for not replying earlier, I have lots of stuff going on right now.

So basically you are describing a scenario where humanity goes digital via WBE and lives inside a utopian virtual reality (either entirely or most of the time) thus solving all or most of our current problems without the need to create FAI and risk utter destruction in the process. Not a bad question to have asked as far as I am concerned - it is a scenario I considered as well.

However, there are many good objections to this virtual reality idea, but I'll just name the most obvious that comes to mind:

If anyone but a FAI is running this utopia, it seems very plausible that it would result in a dystopia, or at best something vaguely resembling our current world. Without some kind of superior intelligence monitoring the virtual reality, mere humans would have to make decisions about how Utopiaworld is run. This human fiddling makes it vulnerable to all kinds of human stupidities. Now, if the timeline this "solution" is supposed to work approaches near-infinity, then at some point collapse due to human error seems inevitable. Moreover, a virtual world opens up possibilities (and problems!) we can hardly imagine right now - possibilities and problems vast enough, that a human mind may simply not be adequate in order to solve and control them.

So unless you are willing to change the human mind in order to adapt it to a utopian virtual reality, it seems to be no viable option. Why mandatory brain-rewiring before uploading a human mind into a virtual reality run by human-like agents would be grotesque I won't elaborate further.

I guess what my objection really comes down to is that humans are stupid, and thus we cannot be trusted to do anything right in the long run. Or at least we certainly can't do things right enough to build a lasting and well-operating utopia. For that to work we would need extensive brain enhancements and if you go down this path, I think you run into the exact same problems I described already:

We humans are self-obsessed shits and bolting on a rationality module won't change that. It would make us better at getting the things we want, but the things humans seem to want most of the time (judged by our current actions) are not working utopias but things typical for self-obsessed evolved agents: Power, social status and sex. Could this change by virtue of becoming more rational and intelligent? There's a good chance it would. People would certainly think more (and more efficiently) about the future once they are smarter, but if you make humans sufficiently smart to produce something like a virtual utopia, you have already taken the first step to creating the uFAI you hoped to avoid in the first place. Humans would want to become even smarter than they (and all those other enhanced humans) already are, thus starting a race to the top. The result may be something terrible that has been endlessly envisioned throughout the ages: Powerful gods with the petty goals of humans.

So my current conclusion after thinking about the virtual reality scenario is that it's not a viable long-term solution for humanity. That is not to say that I've extensively studied and thought about this option, but I think the objection I detailed is pretty convincing.

In a sentence: We're not smart enough to directly construct a utopia (whether real or virtual) and if we actually made ourselves smart enough to do that, we'd probably raise the odds for uFAI. So risk-wise we would be worse off than trying to build a FAI from scratch, which would be more competent at solving our problems than humans could ever be (whether intelligence-enhanced or not).

Also, I realize that a virtual reality is a great way to absolve humanity of resource scarcity and restricted bodily abilities/requirements, but virtually all other problems (moral, social etc.) are left rather untouched by the virtual reality solution. A virtual reality is not automatically a utopia. In a way it would be like living in a world where everyone is immortal and filthy rich - everyone has huge power over material (or as in our case virtual) "things", yet this doesn't solve all your social or personal problems.

Thank you for taking the time to write out such a thoughtful reply. I will be taking the time to return the favor shortly.

EDIT: Here's the (long) reply:

If anyone but a FAI is running this utopia

What does it mean to run a utopia? In order to run something, one must make decisions. What sort of decisions would this person or FAI be making? I realize it's hard to predict what exact scenario it would be like, but we can speculate all we like (and then see which one's sound most realistic/probable). Also, who said that anyONE would be "running" this virtual reality? It could be democratic. Also, who said that utopias have to be with other real people/uploads rather than with non-conscious programs (like in a video game or a dream)? I can understand people's desire to be in "reality" with "real people". But this wouldn't be necessary for video-game type virtual reality.

Now, if the timeline this "solution" is supposed to work approaches near-infinity, then at some point collapse due to human error seems inevitable

I think it is probably less inevitable than the heat-death of the universe. I think that the transition to hardware would permit people to survive on substrates that could exist in space with no need to inhabit a planet. There is no longer a need for food, only electricity (which would be easily available in the form of solar energy). Spreading this substrate in every direction reduce the risk of the collapse of "society".

Why mandatory brain-rewiring before uploading a human mind into a virtual reality run by human-like agents would be grotesque I won't elaborate further.

It won't necessarily be mandatory to rewire oneself to be smarter, kinder, more emotionally stable. But, if one had the petty desires for power, sex, and status (as you claim), then they would willingly choose to rewire themselves (or risk being left behind).

We humans are self-obsessed shits and bolting on a rationality module won't change that. It would make us better at getting the things we want, but the things humans seem to want most of the time (judged by our current actions) are not working utopias but things typical for self-obsessed evolved agents: Power, social status and sex. Could this change by virtue of becoming more rational and intelligent? There's a good chance it would

Today, power is measured in terms of wealth or influence. Wealth seems like it would cease to be a relevant factor as economics is dependent on scarcity (have you ever had to buy air?) and in an age in which everything is digital, the only limitation is computational capacity.

Although this is hardly certain, I hypothesize that ("actual") sex would cease to be a relevant motivator of uploads. Sex in a virtual reality would be free, clean, and offer the user the ability to simulate situations that wouldn't be available to them in real life.

Status is usually sought today in order to have sex (see above) and by means of acquiring wealth (see above).

Personally, I believe that once we become uploads, the chemical imbalances and irrational beliefs that drive our behavior (for evolutionary purposes) will dissipate and we will be infinitely happier than we have ever been.

Powerful gods with the petty goals of humans.

Agreed that it is frightening. Nice way of putting it.

So risk-wise we would be worse off than trying to build a FAI from scratch, which would be more competent at solving our problems than humans could ever be (whether intelligence-enhanced or not).

This is the key question. What is riskier? I acknowledge that P(utopia|FAI+WBE) > P (utopia|WBE). But, I don't acknowledge that P(utopia|AGI+WBE) > P (utopia|WBE).

Also, I realize that a virtual reality is a great way to absolve humanity of resource scarcity and restricted bodily abilities/requirements, but virtually all other problems (moral, social etc.) are left rather untouched by the virtual reality solution

I believe these problems are caused by scarcity problems (scarcity of intelligence, money, access to quality education). And as I've pointed out earlier, I think that seeking sex, power, and status will disappear.

A virtual reality is not automatically a utopia. In a way it would be like living in a world where everyone is immortal and filthy rich - everyone has huge power over material (or as in our case virtual) "things", yet this doesn't solve all your social or personal problems.

Personal problems are caused by unhappiness, craving, addiction, etc. These can all be traced back to brain states. These brain states could be "fixed" (voluntarily) by altering the digital settings of the digital neurochemical levels. (though I believe that we will have a much better idea of how to alter the brain than simply altering chemical levels. The current paradigm in neuroscience has a hammer (drugs) and so it tends to look at all the problems as nails).

So tell me, where does a German psychologist go nowadays to achieve the biggest possible positive impact in the field of friendly AI?

I was/am in a similar position as you are and I've come to the following, somewhat depressing conclusion: Quit studying psychology, earn a lot of money, then donate to FAI-projects is the best you can do if you aren't a math genius. (I don't follow my own advice because all fields related to money-making really turn me off. And because I'm a selfish hypocrite, of course.)

You also seem to overestimate the usefulness of academia in general. Most profs are driven by status, not by utilitarian ideals which makes it somewhat hard to pursue useful research.

If you really want to do the most good you can possible do, you should definitely read this article. ( Just replace "development aid" with "FAI".) I also have some slides from GivingWhatWeCan which I could send you per email.

Oh, and have you read this post?

Quit studying psychology, earn a lot of money, then donate to FAI-projects is the best you can do if you aren't a math genius.

I'm really unsure why this is such a common conclusion. Maybe I'm overerestimating the bar for "math genius", but there is some absurdly high amount of unexplored ground in areas relating to whole brain emulations and brain-computer interfaces. Neither of these are particularly math-heavy, and it's likely that at there are parts of the problem that just require engineering or lab experience. My current estimate is that the more people working on WBE, the better. Also: the larger the fraction of the WBE community that cares about existential risks, the better.

(There's also lots of unexplored ground in machine learning, but having people working on ML is probably less good than having people work on WBE, and also ML is significantly more math-intensive.)

On a separate note, I don't think the people at SingInst, for instance, are so good at math that there is a vanishingly small probability of being better than them, even in fields where math is the determining factor (I don't mean this as an attack on SingInst; the people from there that I've met all seem to be quite reasonable and capable people, for both qualitative and quantitative reasoning. But I don't think that SingInst has been so optimized for math ability that no one else should even bother trying to contribute mathematically.)

I agree with this. Look at the list of SIAI's publications - not all of them would have required a math genius to write.

I think that some of the most important papers relating to the Singularity so far have been (in no particular order) Eliezer's CEV proposal, Omohundro's Basic AI Drives, Robin Hanson's If uploads come first, Carl Shulman's Whole brain emulations and the evolution of superorganisms, Eliezer's Artificial Intelligence as a Positive and Negative Factor in Global Risk, Anders Sandberg's and Nick Bostrom's Whole brain emulation roadmap, lukeprog and Louie Helm's The Singularity and Machine Ethics, and the intelligence explosion paper that lukeprog and Anna Salamon are writing right now. I'd also like to imagine that my two draft papers might have some value. None of those would have required a math genius for a writer - in fact, the WBE roadmap is probably the only one that required any math knowledge at all. Of possible future directions, many of the Singularity and Machine Ethics proposals can be done without being a math genius as well.

Then there are various other useful paths that are less directly associated with research. Popular writing (the career path that I'm thinking about concentrating on) might inspire countless of people to pursue FAI-related pursuits if done well. (Or turn them away from it, if done badly.)

Note that the "it's better to earn money to fund others to do research" presumption assumes that there are people who can be hired to do research. If everyone who's interested in FAI/Singularity issues and isn't a math genius decides to just earn money instead, that means that the only people who can be hired to do work on FAI/Singularity issues are either math geniuses or folks who'd rather be doing something else but agree to do this because it pays the bills. It'd be better to have math geniuses and genuinely motivated folks who were ready to do the tasks that weren't the math geniuses' comparative advantage.

[-][anonymous]12y30

You also seem to overestimate the usefulness of academia in general. Most profs are driven by status, not by utilitarian ideals which makes it somewhat hard to pursue useful research.

Academia has many worthwhile resources. Giant libraries, laboratory funds, political/industrial connections, lawyers, and etc. The existence of status-driven professors does not make useful research impossible (or even difficult, really) to pursue.

Quitting psychology is not an option I'm afraid. Apart from science and rationality, the human mind is the only topic that ever constantly held my interest. Absolutely everything else seems to bore me to shits in no time, because my mind constantly focuses on what it feels to be "important" and my interest nearly hits ground zero as soon as I discover the relative direct unimportance of certain topics on my personal life.

I also have the same problem as you when it comes to making money: Not only do most ways to make some real money feel like a disgrace, I would probably also be slowly dying inside by doing something I hate.

There are however some ways to make some decent cash as a psychologist if you play your cards right... the title of "Dr.Happy" currently belongs to an Australian doctor, but that could be expected to change in a decade from now.

Rationality may also be something you could build a "pop-science-career" on, if the demand for it increases over time. Some rationality is better than none and raising the sanity-waterline probably wouldn't hurt the future prospects of humanity either. (Not that I would intend to do second-class work in such a field as rationality-education, but since we're still talking money, pop-science is where it's at if I stick to psychology.)

[-][anonymous]12y00

There is a lot of useful work that can be done in psychology, largely because it currently is so full of bogus "theories". You might find this post and web site useful, Theory and Why It's Time Psychology Got One. Note that I don't agree with a lot of what the embodied/ecological psychology school is doing, but it gives an interesting perspective on the neuropsych and cognitive schools mostly posted here.

[This comment is no longer endorsed by its author]Reply

Ah, I see, we are on the same page regarding money...

I also think that raising the sanity-waterline is a quite effective strategy. You describe yourself as an extrovert and say that you like writing. Maybe a "pop-science-career" is really the right path for you.

You also seem to overestimate the usefulness of academia in general.

Even if 99.9% of academia is useless, getting 0.1% of it to contribute usefully to FAI research (by e.g. publishing quality papers on related topics and attracting interest) would be immense.

You're right. The point I wanted to make is that especially in the field of psychology ( and especially in Germany) most people don't work on anything interesting at all. In order to secure an academic position you have to suck up to your superiors and work for a substantial amount of time at their projects. And even after getting tenure you have to apply for grants and only get enough money if your research is accepted by the mainstream.

You are correct about the state of German science-affairs, which is why I might be willing to move to the US if that was needed to pursue my career - although I don't know how different the grant-grabbing structure for psychology really is in America. That country is a disgrace on so many levels nowadays that it really would be a hard call for me to move there, but it still gets the most and best science in psychology (and I think also in general) done. I wouldn't be surprised to find that I could pursue many psychology-paths there that simply wouldn't be available to me in Germany.

I have been convinced of this premise and now I would like to devote my life to that cause.

Beliefs are probabilistic. If your beliefs are strong enough that you want to work for one of the related causes championed by the slightly different ideologies of this group, that can be a perfectly fine thing, and you don't have to pretend your convictions are stronger than they are, and you won't be deemed irrational for your hard work despite a lack of certainty. Recall:

"Many psychological experiments were conducted in the late 1950s and early 1960s in which subjects were asked to predict the outcome of an event that had a random component but yet had base-rate predictability - for example, subjects were asked to predict whether the next card the experiment turned over would be red or blue in a context in which 70% of the cards were blue, but in which the sequence of red and blue cards was totally random.

In such a situation, the strategy that will yield the highest proportion of success is to predict the more common event. For example, if 70% of the cards are blue, then predicting blue on every trial yields a 70% success rate.

What subjects tended to do instead, however, was match probabilities - that is, predict the more probable event with the relative frequency with which it occurred. For example, subjects tended to predict 70% of the time that the blue card would occur and 30% of the time that the red card would occur. Such a strategy yields a 58% success rate, because the subjects are correct 70% of the time when the blue card occurs (which happens with probability .70) and 30% of the time when the red card occurs (which happens with probability .30); .70 .70 + .30 .30 = .58...

Even if subjects think they've come up with a hypothesis, they don't have to actually bet on that prediction in order to test their hypothesis. They can say, "Now if this hypothesis is correct, the next card will be red" - and then just bet on blue...

It is a counterintuitive idea that, given incomplete information, the optimal betting strategy does not resemble a typical sequence of cards.

It is a counterintuitive idea that the optimal strategy is to behave lawfully, even in an environment that has random elements...

In the dilemma of the blue and red cards, our partial knowledge tells us - on each and every round - that the best bet is blue. This advice of our partial knowledge is the same on each and every round. If 30% of the time we go against our partial knowledge and bet on red instead, then we will do worse thereby - because now we're being outright stupid, betting on what we know is the less probable outcome.

If you bet on red every round, you would do as badly as you could possibly do; you would be 100% stupid. If you bet on red 30% of the time, faced with 30% red cards, then you're making yourself 30% stupid. Given incomplete information, the optimal betting strategy does not resemble a typical sequence of cards.

Yes, of course I'm very well aware that beliefs are probabilistic in nature and that FAI is no exception. "Conviction" didn't mean to imply absolute certainty on my part, but merely the recognition that creating a FAI seems very likely achievable in principle (although I'm much less certain about the time-frame) and that a FAI is way more efficient to get what people need and want, in comparison to than a million independent half-assed fixing-attempts for each and every problem we face by not-so-clever human intelligences, myself included.

How your quote relates to what you wrote is rather misty to me. I guess that you imply more knowledge of the different viewpoints in the LW community on my part, than I actually have right now. I interpreted what you were saying as: there are different viewpoints on the topic of whether neuroscience is useful for FAI, but most people here think it's not -> so by analogy I'm following the strategy to keep guessing red cards instead of blue ones (which are much more common) - yet you yourself wouldn't say that strategy would be a waste of time and there's still a lower yet reasonable chance neuroscience could actually contribute to FAI after all?

Is that what you meant to imply? If not, then I have no clue how the quote relates to what you wrote.

OK, I can't think of a clear way to say this so bear with me. There is something that is logically true that goes in two directions, like modus ponens/modus tollens do, and I will first describe an inverted version of my argument. This argument would apply to someone whose thinking is the inverse of yours. Then I will explain why I have never in fact directed that argument towards such people. As the last step, I will flip the argument around so it no longer has implications for the person who is in the inverse position of yours, but instead has implications for you - implications opposite of those it had for the hypothetical inversely situated person.

  1. If you are unsure about what is best, there is still a best action that will vey often resemble strenuous effort towards a single goal. Two related but distinct concepts leading to this conclusion are the expected Value of Information and the principle that "... the optimal strategy is to behave lawfully, even in an environment that has random elements."

    So we see that at less than infinite levels of certainty that something ought to be done, it may still be that that thing ought to be done with full focus.

  2. This is a true principle but can lead to cultishness. It is a bad idea to present this true argument to people because once people act with total effort for a cause, they tend to subconsciously violate the principle of behaving lawfully and conclude their total effort is reasonable only if they believe totally, and then they tell themselves that their actions are justified, and consequently that they have justified total belief.

    This would be bad. Cultishness is a characteristic of human groups, so the argument that people should support the group beyond what they intuitively think would lead to net negative consequences.

    It is also true that actions from internal motivations are generally more fruitful than externally motivated ones, so I would let people decide to give their time and money at their own pace.

  3. I deploy this line of argument because you have already said that you have decided you are motivated to work hard for certain things. Assuming you decided at the optimal level of belief, that level of belief isn't so high, and so you shouldn't feel threatened by doubts, obligated to pretend to near certainty, or similar cultish behaviors.

    Just as the uncertain should commit themselves to working hard - though one shouldn't say so, lest cultishness increase - those working hard should remember that an epistemic state under which their actions are perfectly justified is one of considerable doubt and uncertainty.

    So say, if you will, that until evidence indicates otherwise militates, you have devoted your life to a cause. There is concomitant with that no obligation at any level to say that you have been "convinced" of any premise to an extreme degree.

I have programmed some primitive HTML and CSS once and didn't really like it.

This might seem pedantic, but HTML and CSS are not "programming", they are markup languages for formatting text, images etc. A definition of programming might be using a language like Python (or Javascript, for webpages) to make a computer solve some problem/conduct some calculation, and HTML/CSS don't fit this. So, don't take your experience with those and think that that is what "programming" is like.

This comment doesn't deserve to be voted down, s/he is perfectly right.

I have also tried some javascript and disliked it even more. It drives me nuts to not find an error in my code and the act of finally finding and fixing it doesn't feel as rewarding as it should be to make that kind of work enjoyable to me. I can totally see how people love this kind of work but I simply haven't warmed up to it.

I hardly ever get aggressive but boy does fixing stuff for IE6 get my blood boiling...

If you don't want to just use money, it's not impossible by any means. It's merely really, really hard (Edit: on average).

Steps for doing valuable research on a topic:

1) Learn enough about the topic that you know where the unsolved problems are and what attempts have already been made to solve those problems.

This can be a lot of work, but it's necessary. Not only do you not want to duplicate work that's already been done, but you have to know your subject quite well to know what to do without anyone telling you. Working scientists don't just guess at which research avenues to pursue - knowing the subject well usually tells you where the subject needs to be taken farther. Be careful of fooling yourself into thinking you have a good research topic when you don't actually know that to be the case.

2) Pick an unsolved problem and work really hard on it.

Problems in textbooks are doable because they're presented right next to the tools you need to solve them. Research problems are hard because you have to build your own tools, or pull them out of a mountain of not-quite-right options. It's not unusual for problems that took a year to solve to be doable by students in an hour, once they're in a textbook. There are plenty of tips and tricks that you pick up, like doing rough calculations or special cases before the general problem, but really the only way to make a contribution is to work really hard.

3) Go to step 2.

[-][anonymous]12y20

As Vlad says, see money the unit of caring. tl;dr: Don't worry about what you can directly contribute. The best thing you can do is make as much money as possible doing whatever it is you are best at and donate that to the cause (SIAI, I suppose). Given money, a cause will have no shortage of talent available.

You may want to do a bit of time as a visiting fellow or whatever to confirm that your target organisation is in fact what you think it is. There is a lot at stake if you get it wrong...

Setting up a local rationality scene seems like a good idea independent of your particular goals.

I want to know how humans can acquire and impart agency. If you could find that out it would be wonderful for getting people to get better at doing things, including recruiting FAI teams and acquiring money.

Trying to help set up and evolve a rationalist community in Germany would also be a decent task, but compared to specific research that actually directly aids our goals... I somehow feel it is less than what I could reasonably achieve if I really set my mind to it.

I wouldn't write this option off too quickly. It sounds like you would be good at organizing a LW meetup and you would probably get a better idea of what to do from talking it through with people in that community. The people you surround yourself with makes a huge difference to your outlook. Also, you can always start a meetup and do other things.

PS. The QM sequence doesn't teach you QM. It meanly teaches you how to interpret it. Or at least how not to. It also sounds like you should read some of the free will posts, organized here.

lukeprog's and Louie Helm's The Singularity and Machine Ethics draft suggests some useful directions for people wondering about their career choice. Some of them are psychology-related:

On the other hand, value extrapolation approaches to machine ethics face their own challenges. Which value extrapolation algorithm should be used, and why? (Yudkowsky’s “grown up farther together” provision looks especially vulnerable.) How can one extract a coherent set of values from the complex valuation processes of the human brain, such that this set of values can be extrapolated? Whose values should be extrapolated? How much will values converge upon extrapolation (Sobel 1999; Döring and Andersen 2009)? Is the extrapolation process computationally tractable, and can it be run without doing unacceptable harm? How can extrapolated values be implemented in the goal system of a machine, and how confident can we be that the machine will retain our values during self-improvement?

These are difficult questions that demand investigation by experts in many different fields. Neuroeconomists and other cognitive neuroscientists can continue to uncover how human values are encoded and modified in the brain. Philosophers and mathematicians can develop more sophisticated value extrapolation algorithms, building on the literature concerning reflective equilibrium and “ideal preference” or “full information” theories of value. Economists, neuroscientists, and AI researchers can extend current results in choice modelling (Hess and Daly 2010) and preference elicitation (Domshlak et al. 2011) to extract preferences from human behavior and brain activity. Decision theorists can work to develop a decision theory that is capable of reasoning about decisions and values subsequent to self-modification: a “reflective” decision theory.

[+][anonymous]12y-170