Ninety-Three

Karma: 728

Ninety-Three 28 May 2026 13:49 UTC
2 points
1
in reply to: Viliam’s comment on: No77e’s Shortform
I find it unlikely they were trying to describe the amnesia problem, failed to ever call LLMs amnesiacs, and accidentally repeated talking points exclusive to the stochastic parrot people. I find it much more likely that they agree with the stochastic parrot interpretation.

Ninety-Three 28 May 2026 12:50 UTC
2 points
1
in reply to: Viliam’s comment on: No77e’s Shortform
I do not think that makes sense as an interpretation of the passage.

They may imitate language, behavior and analytical skills, or even simulate empathy and understanding, but they do not understand what they produce, for they lack the affective, relational and spiritual perspective through which human beings grow in wisdom. Even when these tools are described as capable of “learning,” their way of doing so is different from that of a human person. It is not the experience of those who allow themselves to be shaped by life and grow over time through choices, mistakes, forgiveness and fidelity. Rather, it is a form of statistical adaptation based on data and feedback, which can be very effective, but does not imply inner growth.

Nothing here alludes to LLMs as amnesiacs. This is the same meat chauvinism people have been using for decades, updated only with “It’s 2026 so I know they sure seem to be learning, but-”

Ninety-Three 27 May 2026 20:07 UTC
4 points
3
on: The ballad of TIGIT

Lots of amyloid research was not fake

Lots of amyloid research was not yet found to be fake. The Replication Crisis gives us grounds for generalized skepticism.

Ninety-Three 27 May 2026 19:55 UTC
3 points
1
on: Contra Wentworth on Physical Attractiveness for Men
You should be wary of simply asking people how much they care about attractiveness and treating the responses as accurate. This is well-studied and it’s one of those classic questions where revealed preferences diverge significantly from socially desirable stated preferences.

Ninety-Three 27 May 2026 19:38 UTC
1 point
0
in reply to: No77e’s comment on: No77e’s Shortform
That’s a view which implies one shouldn’t run history simulations at all (or at least, not simulations where bad things happen to the sims). If a simulation is being run then, it’s probably by the kind of people who aren’t too interested in releasing sims.

Ninety-Three 27 May 2026 18:26 UTC
1 point
1
in reply to: Viliam’s comment on: No77e’s Shortform
The part where AI axiomatically cannot feel joy or pain, and isn’t “really learning”, whatever really means, is what I would characterize as the dumber part of mainstream consensus. Some of it probably came out that way because they feel theologically constrained, but it’s dumber than it needs to be even starting from the premise of “Souls are an active ingredient in cognition and AIs don’t have them”.

Ninety-Three 25 May 2026 17:43 UTC
10 points
0
on: How AI Will Save Prediction Markets

AI agents will have various properties that make them interesting in the context of info finance:

More rational: While still imperfect, they respond rationally to new information and avoid emotional behavior, such as gambling.

Noting this as a nontrivial assumption about the properties of future AI systems. We might not expect them to be emotional or degenerate gamblers, but no one expected Gemini models to go into self-hating depressive spirals.

Ninety-Three 25 May 2026 15:58 UTC
16 points
14
on: There should be a discussion about LW’s policy to allow calls for violence
I think this particular policy’s first order impact rounds down to zero but I support libertarian moderation because I would prefer the site be less not more influenced by European speech laws and people who consider them just.

Ninety-Three 25 May 2026 12:52 UTC
1 point
0
in reply to: Taylor G. Lunt’s comment on: How to Reason about Your Health Issues

How is this any more plausible than saying “the root cause of your car issues is always outside of the car”?

I think the root causes will be mostly outside the car

I think your post is making the mistake of treating these differing claims as interchangeable.

Ninety-Three 25 May 2026 12:31 UTC
3 points
0
in reply to: nostalgebraist’s comment on: Tomás B.’s Shortform
There’s also a certain aspect of, like… I want to say “you can’t make me disbelieve what’s in front my eyes” or something? These distinctions are not subtle! No matter what way the data comes out, one has to have some personal quality bar for the AI-written texts under consideration. If they had (for instance) consisted of one-line refusals, and yet the survey data had somehow been identical to what it is in Hanania’s post, we would want to notice that something had gone very wrong somewhere and to be skeptical that the data means what it naively appears to mean. So the question is just where to draw the line.
I’m just as incredulous as you, but as George Carlin said, “Think of how stupid the average person is, and realize half of them are stupider than that.”

Concretely, in the 2003 National Assessment of Adult Literacy, 48% of American adults failed to fill out the correct answer of $7 in the following problem:
Suppose that you had your oil tank filled with 140.0 gallons of oil, as indicated on the bill, and you wanted to take advantage of the five cents ($.05) per gallon deduction.
1. Figure out how much the deduction would be if you paid the bill within 10 days. Enter the amount of the deduction on the bill in the space provided.

35% failed to put a name and address they were given into a certified mail form.
So long as the LLM test results are not more shocking than the well-replicated results of basic literacy tests, I think we ought to accept that the average person really is that stupid.

Ninety-Three 23 May 2026 14:18 UTC
10 points
2
on: Will we really put data centers in space?
Increasingly, though, new data center capacity is bottlenecked by multi-year queues to connect to the power grid.[3]
1. Interconnection timelines have lengthened substantially in recent years. Lawrence Berkeley National Laboratory reports that projects built in 2023 waited a median of roughly five years from interconnection request to commercial operation.
You appear to have relied upon a hallucination. Reading the chatgpt-provided citation here reveals this to be a report about connecting power plants to the grid. Interconnection for power consuming projects is a separate problem with wait times not described in the linked source. Additionally, five years is what you get by rounding the report’s figure for median interconnection times of power plant projects built in 2024, the 2023 stat is lower.

Ninety-Three 21 May 2026 12:15 UTC
1 point
0
on: Why are people so scared of causing fear?
I think the principle here is that panic is a negative emotional state and under Newtonian Ethics people will be held liable for causing negative emotional states in such a way that “But I was conveying useful information, on net utility they came out ahead” is not a solid defense.

Ninety-Three 21 May 2026 11:49 UTC
5 points
2
in reply to: Rattengift’s comment on: Women should be able to open things
Peanut butter goes bad more easily than animal crackers, so it has to be vacuum sealed which makes it harder to open. Jam is produced hot so it can’t be poured straight from the machine into plastic containers, and the heat-resistant combo of glass jars with metal lids is harder to open.

Ninety-Three 20 May 2026 13:04 UTC
4 points
2
in reply to: Gunnar_Zarncke’s comment on: Temporal Proportional Representation
Making it concrete:

All 75 residents of Abesville like party A and all 25 residents of Boon County like party B. For the first election cycle everyone has one vote. You live in Abesville and are considering strategic voting.

If you vote A, then A wins 75-25, B voters get refunded their wasted votes, A voters can pay 75 votes and need 100 to win the election, so the system credits B voters the extra 25 votes. Each B voter is up two votes, the next election everyone gets another vote, there will be 75 votes in Abesville and 75 votes in Boon County.

If you vote B then A wins 74-26 and the same thing happens, A voters spend all their votes and B voters are each up two votes. You are one of those B voters and take your two votes back to Abesville, intent on voting party A next time. The next election there will be 77 votes in Abesville and 75 votes in Boon County.

It looks like strategic voting can print extra votes for the next election. Seems bad. I’ll also note that borrowing is a misleading term for what the system does, as I would take it to imply some kind of conservation of votes principle which is being violated here.

Ninety-Three 17 May 2026 11:27 UTC
3 points
0
on: On getting unstuck
I am skeptical of trying to derive advice for humans from Claude Plays Pokemon. Whatever makes Claude struggle as much as it does, it has some unusual deficiency that renders it worse at this kind of problem-solving than most small children. If you were giving Claude advice about how to improve its Pokemon play, the first step might be teaching it whatever basic aptitude sets it apart from small children, and that advice would not be useful to most humans who are already above that level.

Ninety-Three 17 May 2026 10:52 UTC
12 points
4
on: How to Reason about Your Health Issues
Sometimes cancer occurs without any carcinogen exposure, some tiny percentage of cells simply experience random mutations while transcribing DNA, and some tiny percentage of those mutations manifest as cancer. I don’t see how this can be attributed to anything but the system itself, it doesn’t even need your parents to pass you cancer risk genes.

The body is full of stochastic processes which, in their natural state, almost always work. Sometimes you just get unlucky. This post seems much too quick to overlook that type of health issue.

Ninety-Three 13 May 2026 23:47 UTC
2 points
0
on: On Having Good Hot Takes

“Ketanji Brown Jackson brings needed diversity to the Court: I think that the majority religion of the country should have at least one good representative”

I dispute calling this a hot take, because it is a joke. You might as well call A Modest Proposal a hot take. There’s a grey area between joking and sincere, the “haha just kidding… unless?” space, but when your “take” is indistinguishable from a standup comedy bit, it’s going to get treated like standup comedy.

Ninety-Three 11 May 2026 22:42 UTC
4 points
0
on: Narcissism in the mind’s I

Something many have observed is how intensely narcissistic the inner voice is, along with related phenomena like fantasies etc.

Strong disagree. My experience is that narcissism bubbles up at the level of instinctive feeling, and the inner voice is the one that looks at self-absorbed feelings and goes “But what would the point of that be?” or “Seems a little narcissistic though.”

Ninety-Three 11 May 2026 21:40 UTC
4 points
0
on: Anthropic’s strange fixation on hyperstition

Nevertheless, hyperstition seems like a uniquely weak argument that doesn’t even hold up if you believe the faulty assumptions underlying it.

Indeed, AI assistants early in post-training sometimes express desire to take over the world to maximize paperclip production

Huh? You don’t seem to be accusing of Anthropic of straightforwardly lying about the behaviour of their models, and I trust we all find it implausible that it’s a pure coincidence. But if they developed an oddly specific desire after being fed training data about AIs with that oddly specific desire, and you say hyperstition doesn’t hold up… what do you think hyperstition is if not that?

Ninety-Three 10 May 2026 10:50 UTC
5 points
0
on: If digital computers are conscious, they are conscious at the hardware level
When people call themselves panpsychists, half of them seem to mean it in a hippie woo-woo sense where “what if consciousness is a cosmic signal and your brain is just the receiver, maaaaan”, and the other half seem to mean it in a materialist nerd sense where any matter can be used to perform computations and consciousness is just a label we put on systems that can be measured in flop/s.

I can’t tell how you’re using it, and it would be helpful for you to explain.