gjm

Hi. I'm Gareth McCaughan. I've been a consistent reader and occasional commenter since the Overcoming Bias days. My LW username is "gjm" (not "Gjm" despite the wiki software's preference for that capitalization). Elsewehere I generally go by one of "g", "gjm", or "gjm11". The URL listed here is for my website and blog, neither of which has been substantially updated for several years. I live near Cambridge (UK) and work for Hewlett-Packard (who acquired the company that acquired what remained of the small company I used to work for, after they were acquired by someone else). My business cards say "mathematician" but in practice my work is a mixture of simulation, data analysis, algorithm design, software development, problem-solving, and whatever random engineering no one else is doing. I am married and have a daughter born in mid-2006. The best way to contact me is by email: firstname dot lastname at pobox dot com. I am happy to be emailed out of the blue by interesting people. If you are an LW regular you are probably an interesting person in the relevant sense even if you think you aren't.

If you're wondering why some of my very old posts and comments are at surprisingly negative scores, it's because for some time I was the favourite target of old-LW's resident neoreactionary troll, sockpuppeteer and mass-downvoter.

Wiki Contributions

Comments

I make no claim to speak for anyone who isn't me, but I agree with your analysis. I would say similar things about e.g. ESP and miracles and the like.

I think calling anything I did "feature engineering" is pretty generous :-). (I haven't checked whether the model still likes FGP without the unprincipled feature-tweaking I did. It might.)

I didn't actually submit a PvP entry. I assume you used my PvE one, but it wasn't intended for PvP use and I am in no way surprised that it came last. I don't particularly object to its having been entered in the PvP tournament, but maybe there should be a note explaining that it was never meant for that?

A quick try suggests that it's working now. I haven't tested thoroughly.

I don't think people who refer to extinction threats from AI as "science-fictional" are making an implicit argument along the lines of "there have been science fiction stories about this, therefore it will not happen".

Their argument is more like "there is no clear sign of this happening in the immediate future, nor an obvious path by which it can happen using only well understood and verified phenomena; so far things like it are found only in science fiction and not in reality, and for it to become real will require other things that so far are found only in science fiction -- e.g., nanotechnology, "backdoors" in the human mind -- to become real first; so if it seems like an imminent threat to you, the most likely explanation is that you are taking science fiction as a reliable guide to future reality, and you shouldn't".

That may not be a good argument! But it's an argument that isn't weakened much by giving examples of things that science fiction predicted and are now either real or imminent.

It is weakened a bit by that, because "X is found, so far, only in science fiction" is less evidence for "we should not worry too much about X" if things found only in science fiction very commonly become reality. But there are plenty of common science fiction tropes that fairly clearly aren't close to happening in the near future -- e.g., faster-than-light spacecraft, teleportation, antigravity -- that if someone finds something credible largely because they've encountered it in SF, then they're probably making a mistake.

Reformatted in what I hope is a sufficiently helpful way.

When I select a definitely non-unique string (e.g., an instance of "amet") and move my mouse over to the smiley face, before the selection is cancelled and the smiley face disappears there is a flash of a larger tooltippy thing which I am guessing contains the warning about non-uniqueness that you describe.

But when I select, e.g., the whole of the first paragraph, or just "Lorem ipsum dolor sit amet", that doesn't happen (or if it does it's too quick for me to see).

So, if I'm understanding what you mean by "unique string" correctly (i.e., a string that occurs only once in the comment), I see different behaviours depending on whether my selection is unique or not, and I get the shrinking/vanishing selection in both cases.

(Which suggests that whatever's going on, it probably isn't of the form "selections thought to be unique are OK, selections thought to be non-unique misbehave", since the unique/non-unique division is visible within the class of selections that vanish.)

Here are some for me (Firefox 113.0.2, Ubuntu 22.04).

  • In some cases the selection (along with the smiley) completely vanishes on mouseover:
    • Paragraph 1: "Lorem ipsum dolor sit amet"
    • Paragraph 1: the whole thing
    • Paragraph 5: the whole thing
  • In some cases, the selection shrinks (and the smiley moves to where it would have been had I made the smaller selection) on mouseover:
    • Paragraph 2: the whole thing
      • Selection shrinks to end after "ali" in the word "aliquam"
    • Paragraph 3: the whole thing
      • Selection shrinks to end after "faucib" in the word "faucibus"
    • Paragraph 3: "Integer dictum tincidunt risus quis varius. Vestibulum erat dui, gravida et commodo"
      • Selection shrinks to "Integer dictum tincidunt risus quis varius. Vestibulum erat "
    • Paragraph 3: "Integer dictum tincidunt risus quis varius. Vestibulum erat dui, gravida et commodo et, dapibus faucibus tortor. Ut sit amet vulputate ipsum. Morbi at blandit nibh. Sed sagittis erat dui, eget placerat dui tincidunt sit amet. Sed ex diam, auctor ut aliquet sit amet, euismod sit amet nunc"
      • Selection shrinks to end after "faucib" in the word "faucibus"
  • In some cases, the selection remains stable on mouseover:
    • Any of the subselections that get shrunk-to as described above
    • All of paragraphs 1-2
    • All of paragraphs 1-3
    • All of paragraphs 3-4
    • Paragraphs 1-2: "Fusce sagittis elit tellus, ultrices maximus velit ultrices eu. Mauris fermentum ipsum vel sagittis dignissim. Sed vitae sem quis dui laoreet consectetur. Cras vel est quis velit imperdiet dignissim nec non metus. Morbi at ligula dolor. [paragraph break] Class aptent taciti sociosqu ad litora torquent per conubia nostra, per inceptos himenaeos. Aenean in sem at mauris euismod condimentum vel nec odio. Vivamus congue est non leo condimentum placerat. Cras interdum mauris quam, non elementum neque aliquet in. Pellentesque risus massa, aliquam id ante a, lobortis varius est. Aliquam erat volutpat."

[EDITED per Raemon's request to include all the text rather than using ellipses for brevity, and to put each example in a separate bullet point.]

Do we know precisely what "credible and urgent" actually means here? E.g., I can well imagine that "complaint comes from a high-ranking officer" is enough to qualify as "credible" and "complaint is of behaviour that would be extremely bad if true" is enough to qualify as "urgent". The point is that these may well be terms of art that don't carry any implication very close to "it is likely that the most incendiary bits of what's alleged are true".

Transmission, preservation and accumulation of knowledge (all enabled by language) are at the top of my list of guesses for most important qualitative change from chimps to humans too.

It certainly seems very likely that AIs can be much better at this than us, but it's not obvious to me how big a difference that makes, compared with the difference between doing it at all (like us) and not (like chimps).

(I exaggerate slightly: chimps do do it a bit, because they teach one another things. But it does feel like more of a yes-versus-no difference than AIs versus us. Though that may be because I'm failing to see how transformative AIs' possible superiority in this area will be.)

A lot of the ways in which we hope/fear AIs may be radically better than us seem dependent on having AIs that are designed in a principled way rather than being huge bags of matrices doing no one knows quite what. That does seem like a thing that will happen eventually, but I suspect it won't happen until after there's something significantly smarter than us designing them. For the nearest things to AI we know how to make at the moment, for instance, even we can't in any useful sense read their source code. All our attempts to build AIs that work in comprehensible ways seem to produce results that lag far far behind the huge incomprehensible mudballs. (Maybe some sort of hybrid, with comprehensible GOFAI bits and incomprehensible mudballs working together, will turn out to be the best we can do. In that case, the system could at least read the source code for the comprehensible bits.)

Load More