If it's worth saying, but not worth its own post, then it goes here.

Notes for future OT posters:

1. Please add the 'open_thread' tag.

2. Check if there is an active Open Thread before posting a new one. (Immediately before; refresh the list-of-threads page before posting.)

3. Open Threads should start on Monday, and end on Sunday.

4. Unflag the two options "Notify me of new top level comments on this article" and "

New Comment
58 comments, sorted by Click to highlight new comments since: Today at 11:17 AM

Stanislav Petrov has died May 20 - information from a local blogger, see comment below the post here: http://turchin.livejournal.com/784006.html?view=5709702#t5709702

Ke Jie is playing AlphaGo. He lost the first game last night, and is generally expected to lose the next 2. According to my buddy in Google they aren't planning to do any more big 'vs the strong player' style exhibitions, so tonight is your last chance to see this John Henry story in real time.

This week's LW2.0 update is late, which is entirely my fault. User interviews had led to a number of improvements to the site, and have made me very glad we're doing a slow rollout.

Last weekend was the CFAR Hack Day, where a team added RSS support to LW2.0. That is, we could set up the Yvain account so that it automatically creates linkposts from a particular RSS feed. Given the potential for spam, this seems like a thing that only admins should be able to add to accounts, but it means we have a complete replacement and integration for the Recent on Rationality Blogs sidebar. We also have plans to include some sort of exclusionary feature. (That is, it'd be convenient for Scott if he could add a 'no-lw' tag or something that causes a post to not get automatically linkposted.)

In what kind of two "person" game would a human have the greatest advantage over a computer? Let's make the game entirely self-contained with objective scoring, no manipulation of real-world objects, and no advantage for knowing real world facts. How far away is "go" from such a game?

Arimaa was designed for this purpose and computers are now better at it.

How about Dixit?

That's no two person game and there's likely an advantage to knowing real world facts.

There's a Dixit-meets-Clue game with a two person variant, but at the moment it avoids me. It's a perfect match for the OP question.

Yeah, I forgot the "two person" contraint. But you can stipulate to do it without reference to the real world and still do well.

Yes, good example.

An interesting question.

My first thought was Nomic, but I don't know how viable is it with two players only. Hmm...

A point for LW 2.0: don't be vulnerable to a spam-vomit script attack (e.g. by using posting-rate caps for new accounts).

This came up in our last meeting, as you might imagine. We already had a 10 minute rate limit implemented for posts and links.

I used to buy melatonin from a place called Puritan's Pride. They ship to the UK, and (shipping included) I'd pay about £15 for 720x200µg. But they've stopped selling it in 200µg; their lowest dose is now 1mg.

Does anyone know any other sites that ship low-dose melatonin to the UK for not loads more expensive? (If you don't know that it ships to the UK, I don't mind checking that myself. I know amazon.com doesn't.)

https://www.vitacost.com seems to be an option, it's more expensive but not prohibitively so.


Perhaps check out Piping Rock? They ship worldwide and their prices seem pretty good. I've gotten l-theanine from there before.

Thanks! They were even reasonably painless, I didn't need to make up a password.

Melatonin is widely available in Europe. If you are willing to use websites in other languages there are many options. For example, Netherlands. I have seen French and Italian sites, but I didn't save them. It was recently legalized in Germany.


Can you buy it in liquid form? Then you can dilute it to whatever dose you want.

Good point, this didn't occur to me. I've had liquid in the past and found it less convenient, but this would have been an option if I hadn't found another supply.

Given perfect conduction (uniform surface temperature, bright side and dark side have the same temperature at all times), https://en.wikipedia.org/wiki/Black-body_radiation#Temperature_relation_between_a_planet_and_its_star applies : temperature does not depend on rotation speed. Then T = T_sun sqrt(R_sun/(2D)) ; it is the temperature T that balance incoming radiation P_inc = pi (R_planet^2) (R_sun^2) (T_sun^4)/(D^2) and emitted radiation P_em = 4 pi (R_planet^2) * T^4

Let's suppose no conduction at all. The bright side and the dark side does not exchange heat at all. Let's take two limiting cases : tide-locked planet, and an "infinitely fast" fliping planet.

In the first case, the dark side of the planet is at absolute 0. The bright side of the planet receives the same incoming radiation but emit half its radiation (halved surface) -- change 4 pi to 2 pi in P_em. Its temperature is T_bright = 2^(1/4) T. Average temperature of the planet is (0+T_bright)/2=2^(-3/4) T

In the second case, each side gets half the incoming power from the sun and radiates half the energy (surface halved). Average surface temperature of the planet is the same that the average surface temperature of any side, which is the same temperature that the perfectly conducting planet, T (the .5 from halved incoming power and .5 from halved outgoing radiation cancel each other).

So : rotation raises temperature of a non-perfectly-heat-conducing planet, bringing its surface temperature closer to the perfectly-heat-conducting planet surface temperature.

This is clearly wrong. The bright side hasn't an uniform temperature T.

You’re right, the 2^(-3/4) (and the 2^1/4) is probably quantitatively wrong (unless each side is perfectly heat-conducing but both are isolated from each other. Or if the planet is a coin facing the sun. You know, spherical cows in a vacuum…). But I don’t think that changes the qualitative conclusion, which hold as long as the bright side is hotter but not twice as hotter than the perfectly-heat-conducing planet.

Oh, yes, it does change, of course.

The result, that a faster rotating planet is warmer, is against the Al Gore's theology about Climate Change, formerly known as the Global Warming.

As Scott Alexander said - the scientific community consensus was wrong, I was right.

I am not sure about him, I know I am right here and the "science community" as it is self-proclaimed - is wrong.

A faster rotating planet is warmer.

So consider an extreme: a tide-locked planet in that same orbit. if it begins to rotate, does the average temperature increase or decrease? I don't actually know, and can't think of a reason it would change. The sunrise will now have access to surface area not yet at the inferno equilibrium, so it'll be absorbing heat. But the sunset is doing the opposite: radiating all that heat that's no longer being replenished.

Does it matter if the planetary material absorbs and radiates heat at different rates? Is that even possible?

Edit: extend this idea. Take a tide-locked planet at equilibrium. Magically flip it so the cold side is toward the sun and the hot side toward space. Give it time to come back to equilibrium. Average surface temperature at t and t is the same, right? During this period, does the average rise and then fall back, or does it drop and then rise back?

I think I recall that hot things radiate faster than cold thing absorb heat, which implies it'll overall cool and then come back up after the now-cold side reaches minimum temperature before the now-hot side reaches maximum. which implies that spinning faster makes the average cooler.

I guess there may be some energy effects of the rotation itself, generating a bit of heat from internal differentials.

I guess there may be some energy effects of the rotation itself, generating a bit of heat from internal differentials.

Heat is actually being generated by the tidal effects here on Earth. The mechanical energy gradually becomes heat.

But those are miniscule amounts, irrelevant for our problem as stated.

I don't know the answer but let me write down some thoughts.

If I'm a point on the surface of a planet then I spend the day slowly getting warmer as I absorb sunlight and the night slowly getting cooler as I radiate my heat into space. If the days are longer then the temperature I reach during the day is higher, which means that when the night begins I radiate heat more rapidly into space. But conversely by the end of the night I've cooled down to a lower temperature so I absorb more heat from the sun at the star of the day.

In fact none of this should matter. Can't we say that the space at that distance from the sun has a particular temperature, and both planets are in thermal equilibrium with that space, so they have the same temperature? That's not such a convincing argument, since the space near a star is not a typical thermodynamic system.

What about atmospheres? They should help warm up the (solid) surface of the planet via the greenhouse effect. I guess the faster spinning planet has a thinner atmosphere, because of centrifugal force, so maybe it's colder.

I think the rate of cooling depends on temperature much more than the rate of warming up, because T_sun - T_planet >> T_planet - T_space. So a faster rotating planet should be warmer.

Cool (heh). Good thinking!

Of course.

And this explains a lot. The so called Faint Sun Paradox is then not a problem at all.

Early Earth was much warmer despite of a fainter Sun mostly thanks to its faster rotation. Partly also because of a smaller distance to the Sun back then, but mostly because of its faster rotation.

It's quite elementary if you think about it.

For now, put the atmosphere aside.

The thing is, that the radiation power is proportional to the T^4. And that the peak temperature matters a lot. Don't you agree?

Under the self-indication assumption, and the assumption that A. Immortality is possible and B. There will only be a finite number of observers, does it follow that an observer-moment should consider themselves immortal, since there are infinitely as many observer-moments belonging to immortal observers than to mortal ones?

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

Just saw this quote from an aspiring game developer on Reddit:

I'm confident in my game, but I'm not confident in others being confident in my game

Such a weird position to be in, and yet many people are in it. When creating any kind of art, it might be a bad idea to focus on making something that you like, because it's easy to fool yourself into liking something and overlooking its flaws over time. A better guideline is to make something that others will like. I know this sounds heretical and contrary to the whole idea of art, but it seems more and more true to me.

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

Prediction is intelligence. Why is there not more discussion about stock picks here? Is it low status? Does everyone believe in strong forms of efficient market ?

(edited -- curious where it goes without leading the witness)

1) Prediction is not intelligence. Intelligence is goal-driven action, which makes use of conditional prediction (if I do X, Y is more likely).

2) stocks are a game where the vast majority of us are at an information disadvantage. It only takes a very weak efficient market to make it unbeatable by the layperson.

3) stock picks are useless for reason #2.

4) calling something low-status is low-status around here. You're better off decomposing the reasons something might affect peer perceptions of the poster.

Just to contradict myself, things like http://cs229.stanford.edu/proj2013/TakeuchiLee-ApplyingDeepLearningToEnhanceMomentumTradingStrategiesInStocks.pdf are somewhat interesting in that they concretely show an application of data.

I personally expect that if something like that worked at all, it'd get incorporated into the standard algorithms and the effect will go away (per the mentioned EMH) within a few months of publication (and note that it's a 2013 paper).

by the layperson

There is a great number of things that a layperson cannot do well. On LW the oft-expressed view is that the market is unbeatable by anyone.

My interpretation of the general sentiment is that market prediction, for it to be remarkably fruitful, has long escaped the possibility for a single individual who is not already blessed by a crapload of money.

Market prediction doesn't require money.

Yeah my comment implicitely assumed that one would have wanted to profit from said predictions.
Still I doubt that a fruitful algorithm for prediction can run on a home computer.

Prediction is not intelligence. Good predictions are one of the signs of intelligence.

Why is there not more discussion about stock picks here?

Apply intelligence :-) If you want to extract money out of financial markets, is picking stocks the right place to be?

Even if you don't believe in the efficient market, picking publically traded stocks yourself means that you believe you can win in zero-sum games against professional investors who are supported by huge computer models and research analysts.

If you have inside information about a company that's not known to the professional investors you might make good trades but you are also violating the law and it would be stupid to publically talk about your trades and their justification on a forum like this.

Another way to make money is to bet on effects that are illiquid enough that professional investors aren't interested. If you have found a trade from which $10,000 can be extracted you are also not benefiting from being public about your predictions. A friend for example used to do arbitrage between different bitcoin markets. While that happened to be profitable, it was stupid to talk about it in a public venue like this.

If you have inside information about a company that's not known to the professional investors you might make good trades but you are also violating the law

As an aside, it's worth noting that insider trading isn't illegal per se (in the U.S.). When inside traders are prosecuted it's often on the logic that they've unlawfully stolen information from the company that they work at, rather than that the trading itself is unfair. If you're walking down the street and find a document that someone has dropped then I believe you can just trade on that information - you have no duty to the company who owns the information. (This is not legal advice.)

Insider training isn't just about an obligation to the company in question. Investopedia:

Insider information is a non-public fact regarding the plans or condition of a publicly traded company that could provide a financial advantage when used to buy or sell shares of the company's stock. Insider information is typically gained by someone who is working within or close to a listed company. If a person uses insider information to place trades, he or she can be found guilty of insider trading. Insider trading is illegal when the material information has not been made public and has been traded on. This is because the information gives those having this knowledge an unfair advantage.

There doesn't seems to be a consensus on this. Also Investopedia

A standard used by the Securities and Exchange Commission (SEC) to determine whether someone who receives and acts on insider information (a tippee) is guilty of insider trading. The Dirks Test looks for two criteria

  1. Whether the individual breached the company's trust

  2. Whether the individual did so knowingly

Tippees can be found guilty of insider trading if they know or should know that the tipper has committed a breach of fiduciary duty.

See also Chiarella v. United States:

The Court held that "a duty to disclose under section 10(b) does not arise from the mere possession of nonpublic market information." Chiarella had no "fiduciary relationship" with either company, nor was he an agent of either company, Chiarella had no duty to disclose the privileged information, and he did not receive confidential information from the targeted companies.

I think that maybe the situation is that the SEC wants insider trading to be illegal but the Supreme Court doesn't.

Okay, that seems like there's more room for doing legal insider training.

I'm trying to find the story of a computer therapist that was rated more positively than a real therapist.

If I remember correctly, the computer would basically rephrase and repeat back whatever you said to it, and people actually found this quite helpful.

Anybody know where I can find that? Maybe a link?

You probably mean Eliza and the story of Weizenbaum's secretary.

I really recommend the book Superforecasting by Philip Tetlock and Dan Gardner. It's an interesting look at the art and science of forecasting, and those who repeatedly do it better than others.

I think we need an association for AGI safety.

I'm poking around using some prototyping tools to try and get something to explain my thoughts about what it might do.


Feedback welcome on the content (warning presentation is bad, do not let operations people near the frontend!)

Differences with MIRI?

Pretty much what ChristianKI said. It is a different niche.

Most current orgs focus on their research agenda for beneficial AI. This org would focus more on understanding on what is currently going on in the world and influencing things socially. It wouldn't sponsor any research directly.

MIRI doesn't do outreach and is not an organization where it's easy to become a member.

"We have the equivalent of a dynamic neural network running our government. It’s ethics free and fed by biased alt-right ideology. And, like most opaque AI, it’s largely unaccountable and creates feedback loops and horrendous externalities. The only way to intervene would be to disrupt the training data itself, which seems unlikely, or hope that his strategy is simply ineffective.


"In a prior column, I discussed the notion that Trump behaves like a machine learning algorithm. Well, his path-independent theory of mind fits perfectly into that metaphor. I’d argue that Trump's path independence operates on multiple levels. It's evident at a meta-political level when he takes a stab at sweeping campaign promises that he never intends to fulfill. It's also visible at the micro level, even within a given sentence: In his very strange recent interview with The Economist, for example, he kept attempting to adjust his message to obtain approval from his interviewers. He keeps things vague, and then pokes his way into a given explanation, but leaves himself room to change direction in case he senses disapproval.


This is a good example of the type of comment I would like to be able to downvote. Utterly braindead political clickbait.

The fact that journalists at a mainstream publication use the metaphor of machine learning to explain the actions of the president is noteworthy. Five-years ago you would be hard pressed for a journalist who thinks that his audience would understand machine learning enough to get the metaphor.

I think it's mainly about combining two click-friendly buzzwords in a novel way.

I like the line about changing his responses to gain approval, is a valid training method...