Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Comment author: ChristianKl 11 May 2017 10:12:38AM 2 points [-]

Anyone knows how intelligence works? Given those two interpretations, do they get assigned priors that trickle up?

If we can believe Tetlock, the US intelligence community didn't assign any probabilities to the interpretations. Afterwards they felt like the screwed up and funded projects like the ACE challenge that Tetlock's Good Judgement Project won via IARPA to find out how things should be done.

The problem with probabilities is that that they produce accountability. If you say X is likely to happen and X happens you can pretend that you meant that it happened with 95% probability and if X doesn't happen you can pretend that your prediction was that it was supposed to happen with 60% probability. As a result there are institutional incentives against using probability.

I don't know what the status quo happens to be within the intelligence community but when I look at the current US administration, I doubt that pushing the CIA to use probabilities more for their assessments is high on the list of priorities.

At the same time getting intelligence right in cases like this is important. That's part of the reason why IARPA spending isn't as evil as weapons research and why 80,000 hours is justified in putting IARPA-Program Manager on their job board.

Comment author: buybuydandavis 23 May 2017 01:41:40AM 0 points [-]

Did they even have "Saddam is faking it" as a possibility?

Comment author: buybuydandavis 11 May 2017 09:43:37AM 0 points [-]

The US intercepted communications where Saddam told his units to ensure that they had no chemical weapons that inspectors could find. Of course, that communication didn't happen in English. That communication seems to have been misinterpreted by the US intelligence community as evidence that Saddam is hiding WMDs.

Even in the English given, I can see alternate interpretations. Make sure you destroy any you have. Make sure they can't find any you have.

Anyone knows how intelligence works? Given those two interpretations, do they get assigned priors that trickle up?

My impression from the Saddam days is that the scenario that Saddam destroyed his weapons the best he could, while trying to maintain the appearance that he had them, really hadn't trickled up to general consciousness, if anyone had considered it at all.

People are overly confident in thinking they've covered the possible motivations a person might have had. "He wouldn't have done this unless."

Any actual science out there on this particular effect?

Comment author: gwern 26 February 2017 02:06:32AM 2 points [-]

A number of LWers signed up (it was posted here) for the BGI high-IQ project; I believe they got copies of their data. As for the project as a whole: Hsu has mentioned some of how it went, including in a podcast last year which I believe got transcribed; basically, BGI made some disastrous strategic decisions in trying to develop & use a genome-sequencing competitor to Illumina and the high-IQ project got orphaned in the chaos, and is largely irrelevant now as similar high-IQ samples turned up nothing special (so the original premise, that either enrichment would increase power dramatically or that rare variants would be found, I forget which, turned out to be wrong) and the regular GWASes like SSGAC+UK Biobank have made much progress & rendered their relatively small sample mostly irrelevant. Hsu sounded moderately hopeful that something might still be finished & published, but hu knows.

Comment author: buybuydandavis 26 February 2017 01:17:41PM *  0 points [-]

So how many 150+ IQ samples did the latest studies have access to?

More generally, what's the equivalent general population sample size for the tail sampled high IQ populations?

Article about the Chinese Study and it's linking up with the SMPY study
http://www.nature.com/news/chinese-project-probes-the-genetics-of-genius-1.12985

Comment author: buybuydandavis 25 February 2017 12:14:05PM 0 points [-]

"We have the tools to do this--we could, for instance, sequence a lot of peoples' DNA, give them all IQ tests, and do a genome-wide association study, as a start."

I remember a few years ago the Chinese offering free genomic scans for the sufficiently intelligent. Did anyone sign up for that? Anyone know of how that story turned out? I assume they weren't going to share that info.

Comment author: Oscar_Cunningham 21 February 2017 12:08:29PM 0 points [-]

I think what Unnamed says is the most important observation:

I'd bet on 6. I have information that the market doesn't have, and my information points to 6 as the answer, so the market is underpricing 6 (compared to how it would price 6 if it had all the information).

Any time you have private information that the market doesn't have you should bet to move the market in the direction of your information. The difficult question is how much you should bet.

Comment author: buybuydandavis 25 February 2017 11:55:36AM *  0 points [-]

What information do they have?

This is the general problem of a mixture of experts when all you have are the predictions but not the information on which the predictions are based (at least for the market). I don't think there is a real answer to that until you input more information into the system.

We want P(side | I_them, I_me)

We have P(side|I_them), P(side|I_me)

The latter don't give the former.

Comment author: ChristianKl 21 February 2017 12:49:18PM *  0 points [-]

Any time you have private information that the market doesn't have you should bet to move the market in the direction of your information. The difficult question is how much you should bet.

I don't think that's true. If it would be true I don't think most mutual fund managers would underperform the SAP 500. A mutual manger might get some superficial information about stocks he invests in that take actual research.

Comment author: buybuydandavis 25 February 2017 11:48:57AM 0 points [-]

Mutual fund managers have incentives other than maximizing expectation of price. OPM.

Also, they likely overestimate the value of their privileged information.

Comment author: jimrandomh 22 February 2017 09:56:23PM 0 points [-]

I believe the argument was (or should have been) that immigration increases the GDP of the destination country by more than it decreases the GDP of the origin country, thereby increasing worldwide GDP per capita.

Comment author: buybuydandavis 25 February 2017 11:27:04AM *  0 points [-]

That should be the argument.

But the argument is also made that "GDP goes up" in the context of the immigration debate and nationalism. That is the usual context of the question these days.

From a nationalist perspective of the existing citizens of the country, their GDP may go down when immigrants are allowed into the country, even while the GDP of the aggregate including the new immigrants is higher than the original GDP of the existing citizens.

And of course, the influx of immigrants also affects the distribution of GDP within the existing citizens.

Comment author: Viliam 03 February 2017 09:58:33AM 0 points [-]

I think people in real life are actually doing this kind of thing, but they don't drive in real traffic; instead they go to an empty place, and try e.g. driving between huge empty paper boxes (so that in case of crash nothing serious happens). And I think they don't do it really to learn driving well while drunk, but rather to calibrate themselves about how exactly different levels of being drunk impact their performance.

So, what would be the analogy of the empty paper boxes for LW?

Comment author: buybuydandavis 08 February 2017 11:53:02PM 0 points [-]

LW is the empty place with the empty boxes are the threads with politics in them.

Though driving isn't the best metaphor. LW is the moot court. Or boxing ring. Or dance floor. Takes two to tango.

We could segregate these things for those who don't feel their up to having a political discussion. Tag a thread with Politics. Don't want to talk politics, don't go into the thread.

Comment author: Elo 03 February 2017 05:14:02AM 0 points [-]

If you practice driving while drunk, you will crash. This will not teach you to drive well (especially if you die). If you practice driving sober, you will get better at driving generally, and also in the domain where you drive drunk. There. analogy fixed. Don't argue over analogies.

Comment author: buybuydandavis 03 February 2017 08:56:45AM 1 point [-]

But driving drunk is not exactly the same skill as driving sober, so that practicing drunk can be expected to improve your skills in driving drunk in ways that only driving sober would not.

Comment author: Viliam 02 February 2017 11:08:18AM 12 points [-]

Debates about discussing politics on LW feel to me like this:

"It is a bad idea to drive drunk."

"How can you say that? There is a huge variation among people with regards to (1) driving skill, and (2) alcohol metabolism. We are the best drivers ever, and we are smart enough to stop driving when we notice that our abilities are impaired."

"Uhm, it sounds a bit like Dunning–Kruger in action, but... a little experiment shouldn't hurt... I guess."

"FULL SPEED AHEAD! THE LAST ONE AT THE CROSSING IS A CHICKEN!"

(Loud crash; explosion; sirens wailing in the distance...)

"Uhm, I guess it actually was a bad idea to drive drunk."

Two weeks later.

"Hey guys, let's drink and drive! What could possibly happen? We are the best drivers ever, and smart enough to stop driving when we notice that our abilities are impaired."

"Uhm, you remember what happened the last time?"

"As the smart and rich people say, past performance is not an indicator of future performance. We have learned so much since then, we are the best drivers ever. Also, THE LAST ONE AT THE CROSSING IS A CHICKEN!"

Comment author: buybuydandavis 03 February 2017 04:10:12AM *  1 point [-]

Except by the analogy, we want to be able to drive better, whether drunk or sober. In fact, especially when drunk, because that's where our poor driving causes real disaster elsewhere, as opposed to the relative safety of our test track.

We want to get Less Wrong, except where it would matter the most.

View more: Next