9eB1

Karma: 1,235

9eB1 2 Apr 2023 18:04 UTC
4 points
0
on: Is there a LessWrong-adjacent place to hire freelancers/seek freelance work?
I suppose you could consider the 80,000 hours job board if not adjacent, at least adjacent-adjacent.

9eB1 10 Mar 2023 16:01 UTC
2 points
−2
in reply to: Cedar’s comment on: Use the Nato Alphabet
According to Wikipedia, the rough timeline is that there were several standards in use by military and civil aviation in the 1940s and earlier, including separate standards for Latin America and elsewhere, and these used Zebra in the US and UK. The International Air Transport Association presented a draft in1947 to the standards body International Civil Aviation Organization that was meant to rectify this, but it still contained Zebra. Apparently this alphabet wasn’t good enough, because the ICAO hired a linguistics professor to create a revised alphabet, and it’s this one that first contained Zulu, published in 1951.

There were several subsequent revisions following this, where the words were tested in university laboratories in the US and UK, before the final list was broadly adopted in 1956. Interestingly, according to this document written in 1959, the team that was revising the 1951 list attempted to replace Zulu with Zebra, but found that it didn’t help the intelligibility of the alphabet, so Zulu was kept in the final standard.

As the next step in the modification, three more words were changed, FOOTBALL, UNIFORM, and ZEBRA were selected for insertion in the alphabet and a new series of tapes, training, and test sessions were instituted. As previously mentioned, the intelligibility of a word is seriously affected by the company it keeps, and it is therefore necessary to test a substitution in the context of the entire alphabet. Articulation scores for individual words should be used only as indicators of the word’s performance in a given context.

A study of the comparative scores on the two modifications and the original ICAO alphabet showed that although the six-word modification [including three words + FOOTBALL, UNIFORM and ZEBRA] is superior to the original alphabet, it falls short in performance when compared with the first three-word modification [excluding FOOTBALL, UNIFORM, ZEBRA] -- more confusions were introduced than were removed by the second three-word modification, although these confusions cropped out in unexpected parts of the matrix. The problem is not unlike that of pushing a dent out of a child’s celluloid ball—even a successful push leaves a small dent in another place.

9eB1 15 Jan 2023 15:10 UTC
6 points
0
on: Is this chat GPT rewrite of my post better?
I do find this one generally better, since I think the overall discussion is not that novel and so doesn’t need such long explanation. That said, it did lose a key justification from the other post, that Twitter’s size makes it akin to the public square, which is why this kind of rule is necessary to contemplate.

9eB1 27 Dec 2022 14:57 UTC
2 points
0
in reply to: wolajacy’s comment on: Fractional Resignation
I thought it would be about resigning from the responsibilities of your job that you dislike to work fewer hours on only the parts that interest you more.

9eB1 17 Dec 2022 3:36 UTC
3 points
0
on: How “Discovering Latent Knowledge in Language Models Without Supervision” Fits Into a Broader Alignment Scheme
I wonder whether there would be much pressure for an LLM with the current architecture to represent “truth” vs. “consistent with the worldview of the best generally accurate authors.” If ground-level truth doesn’t provide additional accuracy in predicting next tokens, I think it would be possible that we identify a feature that can be used to discriminate what “generally accurate authors” would say about a heretofore unknown question or phenomenon, but not one that can be used to identify inaccurate sacred cows.

9eB1 11 Dec 2022 20:29 UTC
3 points
5
on: Simulators and Mindcrime
I downvoted this post because I think this genre of “essay written by GPT” has quickly become tiresome. If someone wants to use GPT as a sort of code completion for essay writing, and thus stands behind all the points, edits it for coherency and to ensure that there are deeper thoughts being explained, it’s no big deal. But just reading unedited GPT essays doesn’t feel like a great use of anyone’s time now that the novelty has worn off.

9eB1 7 Sep 2022 4:52 UTC
2 points
0
in reply to: Eric Askelson’s comment on: Turn your flashcards into Art
For language flashcards, you can have the front card be something like “Biblioteca” and the back card “Library [generated picture of a library].” This helps recall even though it’s not part of the prompt—it just seems to stick better in your memory.

9eB1 17 Aug 2022 14:53 UTC
6 points
3
in reply to: Alex Vermillion’s comment on: The Parable of the Boy Who Cried 5% Chance of Wolf
The reason that the villagers didn’t trust the boy that he didn’t have a track record. One reason we don’t trust people who are loudly proclaiming certain kinds of doom is that they don’t have a track record of accurately predicting things (e.g. Heaven’s Gate), and that’s an inherently important aspect of the phenomenon this post is describing. If the child had accurately predicted wolves in the past, real world villagers would have paid attention to a 15% warning.

The post is suggesting that certain kinds of risks have low probability, and the predictors don’t have a track record of success because it’s impossible, but that they have other lines of evidence that can be used to justify their probability estimates. In the case of “nuclear war that hasn’t happened despite the scares” the evidence is events like the Cuban missile crisis or Petrov Day. But in the parable, it isn’t established that the child has good arguments to justify 5% or 15% wolf appearance rates.

9eB1 17 Aug 2022 2:10 UTC
2 points
0
in reply to: lc’s comment on: Using GPT-3 to augment human intelligence
It’s very cool, especially as a side project. If I’d known it was created by someone here I would have been more careful about the tone of my comment.

9eB1 16 Aug 2022 13:46 UTC
2 points
0
in reply to: Kaj_Sotala’s comment on: Using GPT-3 to augment human intelligence
I tried WriteHolo against an idea I had for a blog post and its recommendations were worse than useless. I think for non-fiction writing it’s probably a non-starter. It couldn’t even give me a concise summary of something readily available on Wikipedia (e.g. the availability heuristic), much less suggest creative connections that are at all coherent.

9eB1 17 Apr 2022 19:51 UTC
2 points
in reply to: gjm’s comment on: Modect Englich Cpelling Reformc
Yes, it is intended to be nonstandard, and underscores the misery of the world, as if to say they things have gotten so hopeless they can’t even muster up the energy to add quotation marks.

9eB1 17 Apr 2022 3:11 UTC
2 points
in reply to: gjm’s comment on: Modect Englich Cpelling Reformc
Cormac McCarthy successfully published a fiction book with no capitalization (and some reduction in other punctuation), The Road, to critical acclaim. Of course, it didn’t lead to a groundswell of novels without capitalization.

9eB1 17 Apr 2022 2:33 UTC
3 points
in reply to: shminux’s comment on: Constraining narrow AI in a corporate setting

Being seen as impeding work and being in the way of money making is the quickest way out of the position where you can affect decision making. Erecting prohibitions would do that.

Disagree. A lot of times management would put someone in the position of writing policy as a matter of pure delegation, in which case they’d want a sensible policy that constrains the business in line with that business’ unique risks. Writing too lax of a policy is worse than writing one that has some restrictions that make sense in the context of what they are actually building. Writing good policy will put you in a better position to write policy in the future.

That said, policy about general AI when the company isn’t going to build general AI is not good. It’s better to focus on realistic scenarios.

9eB1 15 Apr 2022 13:54 UTC
2 points
in reply to: Kinrany’s comment on: Lies Told To Children
I didn’t notice the author, nor the tags. My thought when the article asked for a prediction was that they grew up in a Jainist community of some kind, although I realized that Jainists may not have their own television shows, and the probability of someone from LessWrong having grown up in a Jainist community is probably less likely than that they grew up in a cult of some kind. But cults are even less likely to have their own television shows, I surmised, at which point I decided the point about television shows was probably rhetorical. That’s when I decided the entire thing might be rhetorical, so it wasn’t actually worth making a prediction.

9eB1 29 Mar 2022 23:52 UTC
5 points
in reply to: GuySrinivasan’s comment on: Mental nonsense: my anti-insomnia trick
I learned a similar trick from an old LW post. You focus on the static in your visual field. If it starts to resolve into random seeming images, that is the beginning of hypnogagia, and if it starts to resolve into even more concrete imagery you are very close to sleep. Try to keep focusing on it, eventually you will fall asleep. This generally works for me.

9eB1 29 Mar 2022 12:16 UTC
2 points
in reply to: Vitor’s comment on: Humans pretending to be robots pretending to be human

P.S.: sometimes children do parrot their parents to an alarming degree, e.g., about political positions they couldn’t possibly have the context to truly understand.

It’s much better for children to parrot the political positions of their parents than to select randomly from the total space of political opinions. The vast majority of possible-political-opinion-space is unaligned.

9eB1 13 Feb 2022 21:37 UTC
9 points
in reply to: gwern’s comment on: To Change the World
They could also easily just desire different things (“have different utility functions”). This is the basis for gains from trade, and, more germane to this example, political parties.

If Effective Evil thinks the most efficient way to do evil is assaulting people’s eyeballs with dust specks, and I think the most effective way to do evil would be increasing torture, I can use the money they give me to engineer aeroplane dust distribution technology to reduce torture. If they think 1000 specks equals 1 minute of torture, but I think 10e9 specks equals 1 minute of torture, there is a wide latitude for us to make a trade where I reduce 10 minutes of torture in expectation, and they get more than 10,000 specks-in-eyes. Their conception of evil is maximized, and mine is minimized.

9eB1 23 Aug 2021 2:53 UTC
6 points
on: Anatomy of a full school climate strike
It’s hard not to fight the hypothetical because these movements are determined by how committed the protesters are. If we assume infinite commitment of 100% of 12-17 year olds I think this would be very likely to succeed. If we assume realistic levels of commitment this would never happen. So it’s very sensitive as a hypothetical to your assumptions.

9eB1 23 May 2021 8:10 UTC
4 points
in reply to: Ericf’s comment on: The Argument For Spoilers
Seems you are at least missing one if you think telling someone Bill dies at the end of a movie called “Kill Bill” is your last category.

To get overly analytical, you know it’s a possibility Bill dies. In Sixth Sense you may not even consider the possibility what’s-his-name is dead.

9eB1 2 May 2021 10:51 UTC
3 points
on: Arguing from a Gap of Perspective
I think what you’ve described is most closely related to the Overton window. Often it is discussed in more neutral terms on LessWrong, meaning without the certitude of personal opinions from this post.

Searching for Overton window on LessWrong will turn up more references. If you find this concept interesting, you may also enjoy the Politics is the Mind-killer sequence, which is all about changing your mind on political issues, if you haven’t read that.