Kaj_Sotala(Kaj Sotala)

• Kaj_Sotala 22 Sep 2023 15:08 UTC

  5 points

  0

  in reply to: FeepingCreature’s comment on: Me­mory band­width con­straints im­ply economies of scale in AI inference

  If we had 2.5 petabytes of storage, there’d be no reason for the brain to bother!

  I recall reading an anecdote (though don’t remember the source, ironically enough) from someone who said they had an exceptional memory, saying that such a perfect memory gets nightmarish. Everything they saw constantly reminded them of some other thing associated with it. And when they recalled a memory, they didn’t just recall the memory, but they also recalled each time in their life when they had recalled that memory, and also every time they had recalled recalling those memories, and so on.

  I also have a friend whose memory isn’t quite that good, but she says that unpleasant events have an extra impact on her because the memory of them never fades or weakens. She can recall embarrassments and humiliations from decades back with an equal force and vividity as if they happened yesterday.

  Those kinds of anecdotes suggest to me that the issue is not that the brain would in principle have insufficient capacity for storing everything, but that recalling everything would create too much interference and that the median human is more functional if most things are forgotten.

  EDIT: Here is one case study reporting this kind of a thing:

  We know of no other reported case of someone who recalls personal memories over and over again, who is both the warden and the prisoner of her memories, as AJ reports. We took seriously what she told us about her memory. She is dominated by her constant, uncontrollable remembering, finds her remembering both soothing and burdensome, thinks about the past “all the time,” lives as if she has in her mind “a running movie that never stops” [...]

  One way to conceptualize this phenomenon is to see AJ as someone who spends a great deal of time remembering her past and who cannot help but be stimulated by retrieval cues. Normally people do not dwell on their past but they are oriented to the present, the here and now. Yet AJ is bound by recollections of her past. As we have described, recollection of one event from her past links to another and another, with one memory cueing the retrieval of another in a seemingly “unstoppable” manner. [...]

  Like us all, AJ has a rich storehouse of memories latent, awaiting the right cues to invigorate them. The memories are there, seemingly dormant, until the right cue brings them to life. But unlike AJ, most of us would not be able to retrieve what we were doing five years ago from this date. Given a date, AJ somehow goes to the day, then what she was doing, then what she was doing next, and left to her own style of recalling, what she was doing next. Give her an opportunity to recall one event and there is a spreading activation of recollection from one island of memory to the next. Her retrieval mode is open, and her recollections are vast and specific.

• Kaj_Sotala 20 Sep 2023 18:48 UTC

  5 points

  0

  in reply to: jacobjacob’s comment on: Me­mory band­width con­straints im­ply economies of scale in AI inference

  I don’t immediately see the connection in your comment to what I was saying, which implies that I didn’t express my point clearly enough.

  To rephrase: I interpreted FeepingCreature’s comment to suggest that 2.5 petabytes feels implausibly large, and that it to be implausible because based on introspection it doesn’t feel like one’s memory would contain that much information. My comment was meant to suggest that given that we don’t seem to ever run out of memory storage, then we should expect our memory to contain far less information than the brain’s maximum capacity, as there always seems to be more capacity to spare for new information.

• Kaj_Sotala 19 Sep 2023 6:35 UTC

  6 points

  2

  in reply to: FeepingCreature’s comment on: Me­mory band­width con­straints im­ply economies of scale in AI inference

  To me any big number seems plausible, given that AFAIK people don’t seem to have run into upper limits of how much information the human brain can contain—while you do forget some things that don’t get rehearsed, and learning does slow down at old age, there are plenty of people who continue learning things and having a reasonably sharp memory all the way to old age. If there’s any point when the brain “runs out of hard drive space” and becomes unable to store new information, I’m at least not aware of any study that would suggest this.

• Kaj_Sotala 17 Sep 2023 10:37 UTC

  2 points

  1

  in reply to: rvnnt’s comment on: How to talk about rea­sons why AGI might not be near?

  I think that sharing the reasoning in private with a small number of people might somewhat help with the “Alignment people specifically making bad strategic decisions that end up having major costs” cost, but not the others, and even then it would only help a small amount of the people working in alignment rather than the field in general.

• Kaj_Sotala 17 Sep 2023 8:26 UTC

  26 points

  18

  in reply to: sapphire’s comment on: Clos­ing Notes on Non­lin­ear Investigation

  Would you expect to see retrospectives of cases where they did a good job? If an investigation concludes that “X made these accusations about Y but we determined them to be meritless”, then there are good reasons for neither CEA nor X to bring further attention to those accusations by including them in a public retrospective. Or in cases where accusations are determined to have merit, it may still be that the victims don’t want the case to be discussed in public any more than strictly necessary. Or there may be concerns of a libel suit from the wrongdoer, limiting what can be said openly.

[Question] How to talk about rea­sons why AGI might not be near?

• Kaj_Sotala 6 Sep 2023 19:03 UTC

  12 points

  7

  in reply to: Quinn’s comment on: De­fund­ing My Mistake

  I think that (for this thing and many others too), some people are going to mark you down for it and some people are going to mark you up for it. So the relevant question is not “will some people mark me down” but “what kinds of people will mark me down and what kinds of people will mark me up, and which one of those is the group that I care more about”.

• Kaj_Sotala 30 Aug 2023 18:59 UTC

  3 points

  0

  in reply to: Maxwell Peterson’s comment on: Dat­ing Roundup #1: This is Why You’re Single

  You can still send what’s called an “Intro” to people who haven’t Liked you, but there’s a limitation that non-paid members can only see one Intro message at a time (so if you get one and want to see others, you need to either Like or reject the person).

• Kaj_Sotala 17 Aug 2023 12:38 UTC

  2 points

  0

  in reply to: mcnificat’s comment on: Co­her­ence Ther­apy with LLMs—quick demo

  That is a thing that I have seen happen! The normal reply is to remember that ideal parent figures are intentionally unrealistic and portray a standard that isn’t even meant to be humanly attainable. (Though also I think the version in the chatbot comes off as a little too over-the-top loving, so it doesn’t necessarily quite hit the intended ideal either.) But if you feel like you might react to it this way, it might be a good idea to try working through the anxiety in a less directly triggering way and avoid ideal parent chatbots for now.

• Kaj_Sotala 17 Aug 2023 6:44 UTC

  20 points

  4

  on: Ten Thou­sand Years of Solitude

  Note that while I haven’t seen any outside discussion of this article in particular, the consensus among historians seems to be that Guns, Germs and Steel cherry-picks its evidence and misrepresents its overall facts in order to support a thesis Diamond had decided he wanted to present as true (see e.g. 1, 2, 3, 4). As a result, my personal heuristic is to not believe any analysis that Diamond presents, since there’s a significant probability that it’s misleading in ways that I wouldn’t be able to spot.

• Kaj_Sotala 15 Aug 2023 12:40 UTC

  2 points

  0

  in reply to: Richard_Kennaway’s comment on: Co­her­ence Ther­apy with LLMs—quick demo

  The link works for me, even not logged in

  Well that’s bizarre, for me it continues to work when I’m logged in but not when I’m logged out.

• Kaj_Sotala 15 Aug 2023 5:48 UTC

  3 points

  0

  in reply to: green_leaf’s comment on: Co­her­ence Ther­apy with LLMs—quick demo

  Apparently there’s a filter that shadowbans bots based on some secret arbitrary criteria. Well that’s annoying. I guess I’ll try making it into a GPT prompt instead.

• Kaj_Sotala 14 Aug 2023 17:19 UTC

  3 points

  0

  in reply to: Chipmonk’s comment on: Co­her­ence Ther­apy with LLMs—quick demo

  Huh, that’s weird, it works when I’m logged in to the site with my own account but not when I’m logged out, even though it has been set as public. Also doesn’t seem to show up in the site search. Did the staff shadowban it because they don’t want anyone getting an ideal childhood? :D I’ll have to try to figure out what’s going on.

Step­ping down as mod­er­a­tor on LW

• Kaj_Sotala 14 Aug 2023 9:08 UTC

  12 points

  0

  on: Co­her­ence Ther­apy with LLMs—quick demo

  Cool! Various other people have also had good experiences with using other LLMs for therapy or coaching, e.g.:

  • @KatWoods found GPT to be an excellent therapist

  • A couple of people told me they had really good experiences with this character.ai IFS bot

  • Several folks have said that Hey Pi is excellent for emotional issues

  • I found this character.ai Life Coach pretty good

  • I made this Ideal Mother bot to do something called Ideal Parent Figure therapy; character.ai had limited space for customizing it so it’s a bit one-note, but I had some good experiences with it.

• Kaj_Sotala 7 Aug 2023 12:44 UTC

  10 points

  2

  on: Prob­lems with Robin Han­son’s Quillette Ar­ti­cle On AI

  AI-doomers often suggest that their fears arise from special technical calculations. But in fact, their main argument is just the mere logical possibility of a huge sudden AI breakthrough, combined with a suddenly murderous AI inclination. [...]

  However, AI-doomers insist on the logical possibility that such expectations could be wrong. An AI might suddenly and without warning explode in abilities, and just as fast change its priorities to become murderously indifferent to us.

  I feel like both of these could be interpreted less as “strawman” and more as “imprecise wording”. E.g. “an AI might suddenly explode in abilities and change its priorities” sounds like it could be a description of a sharp left turn, where the AI starts acting hostile to humans after undergoing an increase in capabilities. Hanson has a habit of compressing complicated arguments to just a few sentences, which necessarily causes leaving out nuances like “the change in priorities is not due to self-modification of intrinsic goals but due to increased capabilities outstripping its alignment properties and leading to changed instrumental goals”.

• Kaj_Sotala 2 Aug 2023 11:59 UTC

  3 points

  1

  in reply to: Neil ’s comment on: You don’t get to have cool flaws

  Yeah, I agree with those points.

• Kaj_Sotala 2 Aug 2023 9:37 UTC

  2 points

  0

  in reply to: Roman Leventov’s comment on: In Defense of Chat­bot Romance

  I think these are generally reasonable, but that the prevalence of polygamous societies is an indication that the first point is significantly culturally influenced, e.g. Wikipedia:

  Worldwide, different societies variously encourage, accept or outlaw polygamy. In societies which allow or tolerate polygamy, polygyny is the accepted form in the vast majority of cases. According to the Ethnographic Atlas Codebook, of 1,231 societies noted between from 1960 to 1980, 588 had frequent polygyny, 453 had occasional polygyny, 186 were monogamous, and 4 had polyandry^[5] – although more recent research found some form of polyandry in 53 communities, which is more common than previously thought.^[6] In cultures which practice polygamy, its prevalence among that population often correlates with social class and socioeconomic status.^[7] Polygamy (taking the form of polygyny) is most common in a region known as the “polygamy belt” in West Africa and Central Africa, with the countries estimated to have the highest polygamy prevalence in the world being Burkina Faso, Mali, Gambia, Niger and Nigeria.^[8]

• Kaj_Sotala 2 Aug 2023 9:15 UTC

  7 points

  5

  in reply to: Roman Leventov’s comment on: AI ro­man­tic part­ners will harm so­ciety if they go unregulated

  It’s plausible that there could be such an effect, yes. On the other hand, there are also indications that a similar effect has a role in reducing the amount of sexual violence (countries where porn was criminalized saw significant reductions in their rape statistics after legalizing it), helping de-stigmatize various uncommon sexual fetishes and preferences and thus being protective of the mental health of their users, expose people to more information about what kind of sexual activities they might like (and thus possibly make them happier with positive effects on society), etc.

• Kaj_Sotala 2 Aug 2023 8:45 UTC

  14 points

  4

  in reply to: Roman Leventov’s comment on: AI ro­man­tic part­ners will harm so­ciety if they go unregulated

  Yes, porn is also harmful to the society (but pleasurable to the individuals who watch it). Is this a controversial fact?

  For what it’s worth, my association of the “porn is harmful for society” stance is mostly that of a right-wing/​religious conservative/​anti-sex ideological position. Outside those kinds of circles, I’ve seen some concerns about it giving young people misleading impressions of what to expect from sex, but I don’t recall seeing much of a sentiment that it would be an overall negative—neither from the laypeople nor the social scientists/​sexologists who I’ve been exposed to.