[Survey Taken Thread]

By ancient tradition, if you take the survey you may comment saying you have done so here, and people will upvote you and you will get karma.

Let's make these comments a reply to this post. That way we continue the tradition, but keep the discussion a bit cleaner.

Well, perhaps a bit too simple. Consider this. You set your confidence level at 95% and start throwing a coin. You observe 100 tails out of 100. You publish a report saying "the coin has tails on both sides at a 95% confidence level" because that's what you chose during design. Then 99 other researchers repeat your experiment with the same coin, arriving at the same 95%-confidence conclusion. But you would expect to see about 5 reports claiming otherwise! The paradox is resolved when somebody comes up with a trick using a mirror to observe both sides of the coin at once, finally concluding that the coin is two-tailed with a 100% confidence.

What was the mistake?

Scott Says:

There’s a crucial observation that I took for granted in the post but shouldn’t have, so let me now make it explicit. The observation is this:

No system for aggregating preferences whatsoever—neither direct democracy, nor representative democracy, nor eigendemocracy, nor anything else—can possibly deal with the “Nazi Germany problem,” wherein basically an entire society’s value system becomes inverted to the point where evil is good and good evil.

making physical backups of data

Oh boy, is this ever a good example.

I used to work retail, selling and repairing Macs and Mac accessories. When I'd sell someone a computer, I'd tell them — no, beg them — to invest in a backup solution. "I'm not trying to sell you anything!", I'd say. "You don't have to buy your backup device from us — though we'd be glad to sell you one for a decent price — but please, get one somewhere! Set it up — heck, we'll set it up for you — and please... back up! When you come to us after your hard drive has inevitably failed — as all hard drives do eventually, sure as death or taxes — with your life's work on it, you'll be glad you backed up."

And they'd smile, and nod, and come back some time later with a failed hard drive, no backup, and full of outrage that we couldn't magic their data back into existence. And they'd pay absurd amounts of money for data recovery.

Back up your data, people. It's so easy (if you've got a Mac, anyway). The pain of losing months or years of work is really, really, really painful.

That seems high.

Smoking shortens life by about ten years-- but not so much if you stop by age 40. This may imply that if we get decent anti-aging tech, smoking won't be a serious risk. How hard would the tech be to not be bothered by cigarette smoke?

Let's assume someone who didn't stop smoked for 40 years-- two packs a day. That's 40 x 365 x 40 = 584,000 cigarettes. Divide that into 10 years worth of minutes, and it comes out as .9 minutes, assuming I set up the calculations properly.

This story, where they treated and apparently cured someone's cancer, by taking some of his immune system cells, modifying them, and putting them back, looks pretty important.

cancer treatment link

What counts as a causal problem?

A sufficiently good predictor might be able to answer questions of the form "if I do X, what will happen thereafter?" and "if I do Y, what will happen thereafter?" even though what-will-happen-thereafter may be partly caused by doing X or Y.

Is your point that (to take a famous example with which I'm sure you're already very familiar) in a world where the correlation between smoking and lung cancer goes via a genetic feature that makes both happen, if you ask the machine that question it may in effect say "he chose to smoke, therefore he has that genetic quirk, therefore he will get lung cancer"? Surely any prediction device that would be called "intelligent" by anyone less gung-ho than, say, Ray Kurzweil would enable you to ask it questions like "suppose I -- with my current genome -- chose to smoke; then what?" and "suppose I -- with my current genome -- chose not to smoke; then what?".

I do agree that there are important questions a pure predictor can't help much with. For instance, the machine may be as good as you please at predicting the outcome of particle physics experiments, but it may not have (or we may not be able to extract from it in comprehensible form) any theory of what's going on to produce those outcomes.

You're talking about predicting the actions of an intelligent agent.

LeCun is talking about predicting the environment. These are two different concepts.

Not quite, as SquallMage had correctly answered that 27, 33, 39 and 49 were not prime.

I would always find people in aeroplanes less threatening than in trains.

Hadn't noticed that before but now you mention it, I think I have a weaker version of the same intuition.