Thanks. By "smaller miracle," are you referring to the case that Julia's estimates are wrong? Or something else?
Agreed that expecting a high fraction of observers to survive a nuclear war makes anthropic selection a less-appealing explanation. Would be interesting to see the numbers.
I interpret Julia's "I would have put a probability of maybe 25% on a lot of those. But it starts to add up. [...] which kind of throws into question my ability to assign good probabilities to all of these near misses." to mean that the 25% estimates are either wrong or not independent, because there were too many such events, and the multiplied probabilities of being lucky at all of them is just too small.
So there is probably some other explanation... but Julia in the quoted text does not propose a specific alternative. She just says that if there was only one such event in history, then explanation "25% extinction, 75% we got lucky" would be a good explanation of our current state; but now that she knows there were actually many such events, it does not seem like a good explanation anymore.
I've always thought the British style puts an awkward amount of space between a comma or period and the word preceding it. It's even worse if you start using it with American-style double quotes.
Interesting discussion here on blog of the Chicago Manual of Style, which supports the American convention:
But notice how the commas and the period in the example of Chicago style appear consistently right next to the words they follow [...], creating a pleasing uniformity along the baseline. In British style, placement is interrupted by the quotation marks, though
Agreed, but code generation is a more natural fit for a GPT-style language model. GPT-3 and Codex use massive training sets; I would guess that the corpus of human code reviews is not nearly so big.
The fraction of LW readers who are non-cisgender is much more than 1/20,000. Respondents to the 2020 SSC survey were 94.5% cisgender, 2.5% transgender, 3% other.
Thanks, that's very nice to know!
I'm involved in subcultures with even higher proportion of transgendered people, being relatively fluid myself, so it's always nice to find other contexts in which transgendered individuals have a higher representativeness than they have in the general population.
Really appreciated all ~16,000 words. For what it's worth, I agree with what Pablo said above.
The parts about your ideological journey were quite fun to read. Reminded me of HPMOR Harry learning partial transfiguration—maybe in that he finds a way to see category boundaries that others can't.
I'd be interested to hear more about why you think social transition "doesn't seem like a smart move". I read some of your posts on gender categories, and I feel like I'm agreeing for the most part but losing you in the conclusions. I haven't read much about this, so f... (read more)
In all seriousness, I am hoping to make a change bigger than just the one nap. I do procrastinate going to bed, and you've reminded me that the ill-effects of that might be much bigger than what I usually notice (e.g., long-term health problems, or having less mental energy even when I don't notice feeling tired). The first thing I want to try is measuring how much sleep I get. I installed a sleep tracking app (Sleep Cycle) but I haven't yet had enough time to see if it will give me the data I want, or if it will affect my habits.
you might find exercise 3 useful if your issue is sleep procrastination. I had to slowly block alllllll kinds of things incrementally till I stopped screwing up my sleep
I don’t know why this doesn’t happen in real life, beyond a general sense that whatever weighting function we use isn’t perfectly Bayesian and doesn’t fit in the class I would call “reasonable”. I realize this is a weakness of this model and something that needs further study.
I'll take a stab at this.
You've got a prior, P(dog I meet is dangerous) = 0.99. (Maybe in 99 of your last 100 encounters with dogs, you determined that the dog was dangerous.) You've also got a sensation, "dog is wagging its tail," and some associated conditional probabilities. Your b... (read more)
Because you didn't get actually eaten by the other 99 dangerous dogs, just in a situation where you concluded you could have been killed or severely injured had things gone differently. A "near miss". So you have 99 "new misses". And from those near misses, there are common behaviors - maybe all the maneating dogs wagged their tails also. So you generate the conclusion that this (actually friendly) dog is just a moment from eating you, therefore it falls in the class of 'near misses', therefore +1 encounters.
You have another issue that your 'monkey brain' can't really afford to store every encounter as a separate bin. It is compressing.
It's a bit more complex than that and depends on neural architecture details we don't know yet, but I suspect we can and will accidentally make AI systems with trapped priors.
2Rudi C
Doesn’t this model predict people to be way more stupid than reality?
I think the usual solution there would be for Eliezer to set a reserve price (to avoid the problem of the second-lowest bid exceeding his comfortably-willing-to-pay range).
I’m not saying that major governments keep doing this without any good reason, because I’ve already documented that. There’s the trivial case that everyone thinks it’s not ‘safe’ to give the vaccine to children yet. Many also have legitimate concerns about people who are old and in sufficiently poor health, and governments keep imposing various age rules on eligibility citing safety concerns.
Minor point: I think you might've misread this one? The Facebook policy would allow a claim like "it's not safe to give the vaccine to children" because it does ... (read more)
Is the reference to footnote 1 missing?