sarahconstantin's Shortform

sarahconstantin

https://www.update.news/p/the-rationality-gap Stefan Schubert thinks that, contrary to popular belief, humans often respond appropriately to problems. e.g. just because the Y2K bug was a problem, people assumed there would be a catastrophe, but instead, people solved it.
- there is a popular literature, but mostly debunked by the replication crisis, that promotes the idea that humans are irrational
- you can see your own agency but you can't as easily see that other people have minds of their own too, and may see the same problems you see and act to fix them
- media focuses on exposing institutional failures, but institutions are usually quite good.
  - i don't think i quite agree with him here.
    - i do believe there is a trend of institutional decline in the US and Europe
    - i do think that large formal institutions, including governments, have *always* had a tendency to be authoritarian (believing the institution is entitled to tell everyone else what to do) and a tendency to develop "office politics" where careerism or factionalism works at cross-purposes with the institution's stated mission.
  - i do agree that it's never a big media story when a problem is quietly solved. i remember that there was a lot of talk about a looming grain shortage due to the Ukraine war and when I looked up what happened there it turned out...markets worked. more grain was produced elsewhere and got shipped where it needed to go.
https://marginalrevolution.com/marginalrevolution/2026/08/act-like-it-is-science.html Tyler Cowen continues to be annoyingly unworried about AI, in this case the OpenAI cyberattack revelations.
- i think this is a temperamental thing or an aesthetic thing. he would never admit to being "worried" or "panicked" about anything, ever, because he despises "neuroticism." (and considers it unmanly.) so it's not actually an intellectual issue at all, for him. if he thought a problem were real, presumably he would endorse taking precautions. he still wouldn't call that "worrying", but so what?
- and that raises an interesting question, related to the "rationality gap" linked above.
  - my reaction to the hacks is "seems like the AI industry needs to put more attention into cybersecurity and alignment, and to use different tactics than the ones that resoundingly failed."
  - but that also seems to be the nearly unanimous response from everyone in the industry! even the x-risk-skeptics are "merely" saying that basic non-AI cybersecurity needs to be beefed up. there is going to be a substantial response.
  - so is it "worrying" or "not worrying" to believe this is a real problem that may very well be solved precisely because so many people were shocked and called for action? seems like a difference of semantics.
    - i lean towards calling it "worrying". airplanes are very safe, but they are only safe because it is many people's job to worry about safety. if everyone were blithe and blasé about airline safety, we would have a lot more crashes! "business as usual" does not imply the absence of concern or effort or stress. those things are part of business as usual!
  - Cowen equates worrying with shorting stocks. that seems to imply he thinks "worrying" means believing the industry will not solve the problem (and also that the market is underrating this fact). I think that's a pretty unfair framing. The way industry would solve this problem is by lots of people sounding the alarm and saying this sort of thing must never happen again! that does not make those people incorrectly alarmist.
  - also, I don't think it's entirely fair to ask everyone to make a quantitative estimate of the cost of future cyberattacks before they are allowed to say "this sounds bad, man." I don't think I've ever seen forecasting of cyberattack cost, even by security experts! i don't think there's really a "science" of this. it would be more like guessing.
    - and I think Tyler Cowen himself has written about the futility of (pop) Bayesian guessing. putting numbers on your subjective beliefs when those beliefs are not based on much factual knowledge. Asspull numbers don't clarify that much, and they give the illusion of knowledge where it doesn't exist.
      - i never give asspull numbers myself. in my case it's because despite extensive practice on Metaculus and Fatebook, i am not very good at Fermi or probability estimation. yes yes, maybe revealing my weakness would be Justice here, but so few other people actually put numerical estimates on things that I wouldn't really even have a peer group to compare to.
https://www.theargumentmag.com/p/i-dont-need-you-to-believe-im-a-woman Cassie Pritchard says trans rights should not be premised on philosophical debates on the meaning of gender. i think this argument is solid.
https://neurobiology.substack.com/p/the-wrong-gustav-mann lovely history of the origin of perfusion fixation by Andy McKenzie
https://arxiv.org/pdf/2402.01702 "Fluent Dreaming for Language Models", like feature visualization but for language models

Cassie Pritchard says trans rights should not be premised on philosophical debates on the meaning of gender.
I can’t really imagine why someone would bother going through hormone therapy, hair removal, and difficult surgery recoveries if they didn’t think all that work resulted in a genuine and thoroughgoing transformation.

Well, neither can I, but there is a well-documented thing called Factitious disorder imposed on self which sounds similar. I am not saying that it is necessarily the same thing, just that "I can't imagine why someone would do such thing" ... (read more)

sarahconstantin's Shortform

8

sarahconstantin's Shortform

8