All of akvadrako's Comments + Replies

In any case, are you making the claim that if a neural net were able to figure out the rules of the game by examining a few million games, you would accept that it's a universal knowledge creator?

If it could figure out the rules of any game that would be remarkable. That logic would also really help to find bugs in programs or beat the stock market.

Not true. Most financial markets are prediction markets. They seem to be popular.

That may be technically true, but only in a superficial sense. Stocks prices have a very complicated relationship to real world events, except in the very long term. That's very different from markets like which have clear connections to things like who will win elections and objective criteria.

First, I didn't say "stocks", I said "financial markets". Second, all markets, prediction included, have a complicated relationship to real-world events. The markets strongly react to some, weakly react to others, and ignore the great majority of them. I think you're trying to say that financial markets ignore some events you're interested in. That's a fair point, but it also applies to all markets.

You should add a "None of the above" option. If I saw an app with these names, I'd be hard pressed to give it a chance.

You might want to try ; crowd sourced names; pay the winner $100.

Tell this to Vladimir_Golovin. I have just installed the poll, exactly for the names he proposed. I think, the option "none of the above" is useless. Imagine some future parents have a list of potential names for their offspring. Mary, Magdalene, Judith and Margarete. They show you this list and ask you which name would you choose. Would you still demand "none of the above" option? (Applications are somewhat like children.)

Also like username2, I'm happy to hear of others with a view along this direction. A couple years ago I made a brief attempt at starting a modern religion called noendism, with the sole moral of survival. Not necessarily individual survival; on that we may differ.

However since then my core beliefs have evolved a bit and it's not so simple anymore. For one, after extensive research I've convinced myself that personal immortality is practically guaranteed. For another, one of my biggest worries is surviving, but imprisoned in a powerless situation.

Anyway, those details aren't practically relevant for my day to day life; these similar goals all head in the same direction.

Sorry, but the idea that Esperanto is somehow only easy for French speakers is plainly wrong. I don't think you'll find anyone who has learned it and another language who'll disagree.

Actually Esperanto is in the same language family as many Asian ones:

You should better look at the wikipedia page I linked:

Also it's not about being similar to French and I don't know why you think that. I've learned Esperanto and French and didn't notice any similarities. Actually the Chinese were one of the biggest supporters, though that may be trending down.

It would be easy to grow Esperanto quickly. It would require some concerted effort, but there is a solid though small base around the world and there only needs to be some push to make it happen. Becoming ... (read more)

Because (1) the study mentioned in the Grin report was conducted on francophone students and (2) while Esperanto is a proposal for a universal language, its structure and vocabulary are very decidedly European and indeed Romance. It is much more like French than Japanese or Mandarin or Korean, or even Sanskrit. Or, in fact, German. That surprises me. Let's try a little experiment. Go to the Wikipedia page on Esperanto (selected just because it's an obvious thing to select, so you know I'm not cherry-picking) and find the first substantial quantity of Esperanto text. It's this: The very first word (en) has approximately the same spelling, pronunciation and meaning as a French word. This is not a coincidence. The next word doesn't (I think). The next (lokoj) is in fact cognate with French lieux with the same meaning. Next (de): French also has a word "de" with the same spelling and similar pronunciation, and a closely related meaning. Then Ĉinio; corresponding French is Chine, similar spelling, similar pronunciation. Maybe half the words in this passage have close French cousins. The sentence structures are very similar too. The writing system is almost identical -- same repertoire of letters, similar set of accents, more or less the same punctuation. If you took the same text and wrote it in, say, Tamil, it would be very much more different. Easy for whom? What's the actual sequence of events that would lead to it happening? I think we may have different ideas about what constitutes plausibility. I agree it's possible but I'd put the probability well below 1%.

I think you're being too pessimistic about Esperanto:

  • There are about 2 million speakers worldwide [4]. For a language only 100 years old.
  • It was recently added to Duolingo [5], a great resource for learning.
  • The Esperanto wikipedia is ranked #32 in terms of number of articles. [1]
  • It's taught in 69 universities in 24 countries, several offering bachelors or PhD degrees. [7]
  • Prominent people are fluent in Esperanto, like the president of Austria [8]
  • After Britain leaves, only Ireland will speak English in the EU, giving Esperanto an opening. [11]
  • Esperant
... (read more)
I don't see how those numbers, even if correct, mean that I'm being too pessimistic about Esperanto. I didn't deny that some people speak it, or that it's easy to learn. I said I don't see any plausible pathway by which it becomes widely enough used to be a lingua franca. The most interesting of those figures is the one about how many hours it takes to learn various languages. The link you gave doesn't offer any direct support for the startling claim you make (apparently saying that Esperanto is 10x easier to learn than English); rather, it quotes someone else describing a study apparently done by the University of Paderborn's Institute of Pedagogic Cybernetics. (On French students, so part of what this is measuring is similarity to French; that will no doubt be why German is alleged to be harder than English. I remark that Esperanto is more like French than English is -- though probably not more like French than Italian is.) Unfortunately I can't readily track down more information about this (it's cited in an article by Flochon in a book by Guy Gauthier but, at least as quoted in the Grin report, doesn't give any specifics about the study). I would want to know more before believing that the ratio is so very large.

I think you are missing the point. If I have a random variable between 0 and 10, than "random" changes will cause a regression to the mean. Thus, if the current state is bad, say 1, a many "random" changes are likely to be an improvement.

More simply, if our state is bad, we should take more risks.

The model of a scalar between 0 and 10 is bad because it doesn't show the high dimensional nature with many different scenario that real world society has. We have a huge decline in violence in the post-WWII are. The mean of history has a lot more violence as Steven Pinker lays out in "The Better Angels of Our Nature".

The argument rests on that assumption, mostly clearly shown in the quote:

"People who voted for Trump are unrealistically optimists, thinking that civilization is robust, the current state is bad, ..."

If we are stuck in a locally optimal valley, then a high-variance candidate is more likely to push us out of it and into another valley. Whether that's a good idea depends on if our current state is overall good or bad.

Personally I think we should be taking more chances and trying to find a better equilibrium. That means occasionally rocking the boat, but if you never do it you're condemning yourself to stagnation.

No, it depends on whether random changes to our current state are an improvement or aren't. If you would make a change that requires all high level government burocrates to be superforcasters in their domains of expertise, it would likely be a huge improvement and you could speak of the resulting government as being good and the present one as bad. That doesn't mean that randomly breaking things and creating change improves the bad current state. A lot of possible changes lead to WW3 or otherwise end civilisation.

You could publish it as GPL3 or something more restrictive. If someone else has a plugin that has commercial potential, they'll need a more permissive license.

Just post it on github with no effort. If you start getting pull requests or issues logged, you'll have your answer.

It doesn't really matter if it's immediate according to empty individualism. Instead the chance of survival in the branches where you try to die must be much lower than the chance of choosing that world.

You can never make a perfect doomsday device, because all kinds of things could happen to make it fail at the moment or during preparation. Even if it operates immediately.

I think we agree but I was trying to make a bigger point than your reply captures. I doubt that you will even experience the terminal illness assuming there are many more possible futures where you stay healthy and anti-aging science advances than ones where you are miraculously saved at the last minute, by aliens or luck.

That makes the roulette scenario relevant to our experience. Because if you have the conviction to pull the trigger if something doesn't go your way you have the three options I laid out. So most likely you don't even have to try - assuming you are sure you will.

In that case the only outcome the participant can expect to experience, and that they will experience with certainty, is that the gun didn't fire

Yes, that's the point. Every future version of you will of course call themselves "you".

Note that playing quantum roulette successfully depends crucially on the speed with which you can kill yourself. With this I agree, which is why I think the quantum Russian roulette or quantum suicide scenarios are mostly interesting as a thought experiment, as they're intended to be.

Although I don't want to ... (read more)

Yeah. I meant that I don't find the roulette scenario very relevant since I believe that we're much more likely to experience some other scenario where this property of the quantum multiverse becomes relevant, like the terminal illness one I described. Most of us won't play the roulette. Anyways, there's a flaw in lisper's original argument: death is not unavoidable even after wrists have been slit.

Hi lisper,

I found your paper easy to follow and maybe insightful (I'll have to read it more carefully the second time) but like qmotus, I don't understand your reasoning in this thread. I'm assuming MWI is just an interpretation of unitary QM, so makes all the same mathematical predictions as other non-collapse theories. And the roulette story is just one way of looking at it, from the perspective of what I consider my (classical) self and what I call the future.

Since you are not claiming that QIT makes different mathematical predictions than MWI, how can you claim they make different predictions at all?

QIT and MWI don't make any different predictions that are testable in a single classical universe (obviously, because QIT and MWI are just different interpretation of QM, so they both make the same predictions for all observables, namely, the predictions made by QM). QIT and MWI are simply differences in perspective -- the God's eye view (MWI) versus the mortal's-eye-view (QIT). Neither view is "correct", but since I (the thing engaged in this conversation) am a mortal, I choose the mortal's-eye-view as more relevant for day-to-day decision making. But as I keep saying, it's ultimately a matter of personal preference. The problem with quantum roulette is that it takes a prediction made from a God's-eye-view and tries to apply it in a mortal's-eye-view context. Yes, God will be able to see that there is a you that survived the process and went on to live the life of Riley. But whether or not you will be able to see that is a very open question. (God will also be able to see a lot of branches of the multiverse containing your friends and loved ones mourning your untimely death.) Note that playing quantum roulette successfully depends crucially on the speed with which you can kill yourself. Trying to play by slitting your wrists, for example, doesn't work because once you see that your wrists are slit you can't roll that back. So the success of the enterprise depends entirely on killing yourself fast enough that you don't become aware of your imminent and (in the relevant branches of the multiverse) unavoidable death. How fast is fast enough? Well, that is (literally!) the sixty-four-million-dollar question. Unless you have an answer that you are very confident is the correct one, it seems to me like an imprudent risk to take.

I am one of those proponents of stopping all AI research and I will explain why.

(1) Don't stand too close to the cliff. We don't know how AGI will emerge and by the time we are close enough to know, it's probably too late. Either human error or malfeasance will bring us over the edge.

(2) Friendly AGI might be impossible. Computer scientists cannot even predict the behavior of simple programs. The halting problem, a specific type of prediction, is provably impossible in non-trivial code. I doubt we'll even grasp why the first AGI we build works.

Neither of t... (read more)

This sounds like an excellent idea if you can get good number of countries represented. Could you clarify a couple things?

  • Is that €70 per night or for the whole weekend (and with 2 breakfasts)?
  • How many participants can attend?
  • How many are signed up already?
I'll gladly clarify things: * The 70 Euro mainly include a bed in a shared room for two nights, two breakfasts and a part of the conference room rent. We've tried to keep it as cheap as possible to allow everyone to attend. * The conference room fits 40, maybe more. We currently only booked 30 beds but could surely extend that. * Signing up involves transferring money, possibly internationally, and banks are ridiculously slow at executing these transfers. So only one person has successfully signed up by now. Maybe some bitcoin transfers have reached John already, I don't know. Yesterday evening there were 35 people who are definitely interested and 4-10 maybes.
To sign up, I had to wire transfer 70€ for the whole weekend. Concerning breakfast, the post above now says: "The cost is 70 € including accommodation and breakfast."

I'm going to have to pass on this one, but I'd like be interested to see comments about how it went. Hopefully this becomes a regular thing in Amsterdam / Utrecht.

Would this be the first Less Wrong meetup in Amsterdam? Sounds worth a try.

First of all, thanks for running the contest Alex and congrats to the winners.

Random behaviour was obviously a good strategy, but I don't feel it teaches us much about making better bots. If I had suspected so much random behaviour, defecting whenever it's detected would have been a good additional test.