Wiki Contributions


Humans are very reliable agents

I'll add to this: humans make errors all the time, but we have institutions and systems in place to ensure that those errors don't lead to catastrophe, by having additional safeguards beyond just hoping that one person doesn't screw up.

That Alien Message

I'm glad you wrote this comment because I gave up on the post about 75% of the way in. Then I read your comment and went back and finished it, and liked the twist.

Whence the determinant?

det(AB) = det(A)det(B), so the determinant is invariant to any change of basis, not merely unitary ones:

Scott Alexander 2021 predictions: calibration and updating exercise

Does “vaccinated” mean at least one dose? Or “fully” vaccinated?

How should I behave ≥14 days after my first mRNA vaccine dose but before my second dose?

I'm aware that there's an ongoing debate about the tradeoffs of prioritizing first doses for as-yet unvaccinated individuals vs. getting people "fully" vaccinated more quickly.

However, one could very well argue for prioritizing first doses over second doses, even if there were a substantial increase in efficacy provided by that second dose, simply because the marginal improvement of going from 0 doses to 1 is much higher than going from 1 dose to 2.

What I'm asking about is whether there's really any evidence for this increase at all, or whether it's just as safe to begin my air travel 14 days after my first dose as 14 days after my second.

Pseudorandomness contest: prizes, results, and analysis

There is something I found really interesting about Ben’s submission, which is that his probabilities went as high as 88%. Ben’s was an outlier among the best entries in this sense: most of the entries that did best had a maximum probability in the 60-80 percent range. This confused me: could you really be 88% sure that a string was random, when you know you’re assigning something more like 60-65% to a typical random string? Can a string look really random (rather than merely random)? This sort of seems like an oxymoron: random strings don’t really stand out in any way sort of by definition.

In principle, at least, the explanation would be straightforward: suppose X is a measure of how many tests of randomness (i.e. tests of the presence of some feature which we expect to more commonly occur in real strings than fake ones) a string passes, and we assign our probability monotonically from X.

Your typical fake string fails many tests of randomness, whereas your typical real string fails few. Yet, there is still variation in X within each category. Hence some real strings will pass nearly all the tests, and get an unusually large probability of being real.

Some criticisms of polling

Can you clarify the meaning of the sentence, "late polls had given H Clinton a win probability over 80%"?

Polls don't give win probabilities, they give a sample of people's opinions, from an average of which you can perhaps derive a win probability in conjunction with a statistical model such as FiveThirtyEight's (which gave Clinton a 71.4% chance on Election Day).

What happens to variance as neural network training is scaled? What does it imply about "lottery tickets"?
Answer by dsjJul 28, 20209

One assumption that I think might be implicit in your question is that the number of lottery tickets is linear with model size. But it seems plausible to me that it’s exponential in network depth.

Has the effectiveness of fever screening declined?

But in the context of the question as to whether it might exert selective pressure, even a ~10% reduction in R₀ would be quite relevant, when compounded over one or two dozen generations of the virus.

Has the effectiveness of fever screening declined?

I don't understand why you say it was never effective. Certainly some infected have no symptoms, or no symptoms yet, or different symptoms, but fever is among the most common symptoms, and so it does catch a significant fraction of active infections.

Load More