This article deserves dedicated post status. He comes in more gently than I would at the end, but the tl;dr that needs to be hammered into peoples’ heads: Hacker News is not a neutral forum for technical discussion. It is an advertising and propaganda operation administered to shape and condition technical discussions.
https://reason.com/2026/09/22/the-ai-safety-movement-is-making-ai-less-safe/
The folks at Reason continue to misidentify as libertarians, not realizing they’re advocating for the shambling corpse of what was already an unhealthy movement
Damn, who knew this would yield material for my boy-band cover/parody project. Foom flop, doo wop wop wop…
To me, the difference is that the calculator was never really imposed upon you. It was obviously useful in certain situations (vanishingly few people are interested in calculating a sine or a square root by hand, practically none in doing so on a regular basis). Many of these AI tools are, in fact, being imposed through top-down mandates and interface dark patterns. That sets up the sort of adversarial environment that leads to learned helplessness over time.
Even from beyond, he’s still sticking his fingers into our hearts ❤️
I think it’s a bit under-covered that the weird SF/Bay Area group-house scene has to be symbiotic with the housing shortage in the region. Even in my old hometown of Hayward, all the older 1950s-70s tract-home and mixed-density neighborhoods tend to have cars parked everywhere along the streets where a car could possibly be parked, because all these places that were designed as single-family homes are being used as multi-tenant rentals. I guess one underrated reason to support YIMBY is to strangle your local incipient cults.
I’ve seen “cognitive surrender” bandied about fairly often in various places online throughout the year, but I also keep wondering if the already well-defined psychology term “learned helplessness” doesn’t cover the underlying concept just as well.
Bonus: A galaxy brain take: maybe since Caroline has already messed up so massively, that means she actually has the best experience!
Yeah, this Mohamed Atta guy is a great airline pilot, he just had a botched landing that one time
Well, I suppose I should thank Mr. Aaronson for conveying to us what it feels like to be a precisely targeted demographic
“YOU MUST TRY THE HOSTAGE SPECIAL!”
I guess I don’t hang out with enough photographers :)
Murdoch having a lucid moment and realizing the sloppers can generate vastly more slop than his people can?
(complete with a reference to “EV” (expected value))
In this day and age, if I see someone talking about major life decisions in terms of +/-EV, I turn on my heel and walk away rapidly. It was kind of cute when this jargon was entering wide use during the online poker bubble, but that was 20 years ago.
What the actual fuck?
This third category is the Patrick Bateman types who would have mainly been confined to finance at the turn of the century. They probably picked San Francisco over New York in part because they were worried the New Yorkers might make them wear a suit.
There are two specific aspects of Rationalism that cause this mysterious blind spot. First, they believe that superintelligence is inevitable, which conveniently removes responsibility from the people actively trying to bring it about. Second, they treat everyone outside the Rationalist sphere as NPCs who dutifully follow the directions on their script.
I’ve been thinking more about this lately, and my metamodel of the rationalist world model (if I may get away with such a pretense) is that they subtly believe that the future is already something crystallized, even if they don’t consciously realize they hold this belief. It is not really a probabilistic philosophy, no matter how much they assert that. Intelligence, to them, is a function that can more and more precisely apprehend the future as it escalates to greater values along the g-scale, and the singularity will manifest an intelligence that can do so with practically absolute precision (hence its prophesied ability to outcompete and replace humans).
When I’m talking to people who haven’t encountered this stuff as closely, I half-jokingly describe the reflexive assignment of probability estimates as a devotional practice akin to prayer or meditation… but it really is! It’s a subtle invocation of the AGI/ASI god, in that I make this imprecise estimate now, to demonstrate my belief in a precisely quantifiable future; but come the singularity, God will relieve my feeble human brain from having to commit this lamentable imprecision. In that reckoning, those who are not even willing to venture an imprecise esimate must have no intelligence worth noting.
The OpenStax textbook covers the development of the general equations of the various conic sections. If the Stewart book covers that in such detail, it must be jammed into the appendices or exercises someplace. I’ll have to find the time to go over that, at least; I still don’t have a great intuition for hyperbolas.
Levi-Civita_symbol and the Einstein summation convention
Yes, this looks like what I need to understand the negative j term in the cross product formula. Cambridge looks to have the 3rd edition of Boas’ book up for download, so I have a copy now. We’ll see how well I comprehend it, I feel like I’m missing the concept of index permutation to really understand the Wikipedia article.
I don’t know how much traffic we’d get if we tried to cultivate a math discussion here. I did make an account on Mathstodon too and perhaps I should spend more time there. Tutoring at this level is sparse, at least at a community college. There’s one tutor for my class available via Zoom, and the instructor himself is apparently teaching 5 classes when he’s only under contract for three. If I signed up for 5 classes as a student, I’m sure I’d be setting off all sorts of alarm bells in the registration department, if they even let me go through with it.
Ah, “Kurzweilian h+,” there’s his marketing tagline. Too bad importing Kurzweil also brings in all his bullshit extrapolation about exponential curves that all the other weird shit is based on. You’re never going to shout down the hyperbolic lunatics by adopting the core thesis of their hyperbole.
Thanks. So far, the questions I missed on the first exam were mainly due to sign errors when calculating determinants for cross-products, so that’s a simple enough fix I suppose. I’ve seen some of the material before, as halfway through the last time I took Calc 3 was when the Covid lockdowns hit. Even though classes are back in-person, this current instructor is still presenting most of the material in static notes and lecture videos online, so it’s still a tough adaptation for me. We’re using the James Stewart text that is pretty common everywhere, though thankfully the instructor has set it up so I don’t have to pay for ebook access yet again.
I picked up the famous Spivak text a few years ago, and am only just now working up the courage to crack it open. I’ve had a few years’ gap in between classes, so reviewing the earlier material with more proofs involved is probably a worthwhile exercise in any event. I’m given to understand that particular text is something of an on-ramp to real analysis. I’ve been considering trying for a math minor once I transfer for engineering, and it seems like real and/or numerical analysis would be two of the prime upper-division classes for that.
Basically segyges feels like controlled opposition, just not by EAs, perhaps other people who don’t mind that he never brings up the pervasive eugenics angle? He certainly posts (including in LessWrong) like it’s his day job.
Maybe, but I’m also willing to buy that Gyges (and the other tech-employed wannabe-calm AI centrists populating Bluesky) genuinely believes in the broad concept of “AI” as the necessary next stage of techno-progressivism. He and his fellow travelers are just positioning themselves as the sane, sober, starched-shirt Presbyterian sort of believer, as opposed to the tongue-speaking, snake-handling Pentacostals that are the hardcore rationalists.
It’s important to understand that the eugenics aspects are never going to be a particularly high concern for people like Gyges. They’ve comfortably made it past the selection filters in education and employment that are relics of the eugenics movement, so they think IQ-supremacy and neo-phrenology can’t hurt them anymore. It’s when this attitude curdles into fear that someone might revoke the status conferred by passage through such selection filters that you get a DHH, Shuttleworth, etc. I think that’s a prominent root cause as to why I, and many of us here, can’t trust a Gyges type. They may differ with the fascists right now, but they also have no innate defense against ethical freefall.
Even though things are bad now, the conservative movement in America is likely to be severely fractured post-Trump. I don’t think there’ll be a unified movement there for Yud to flop into. I think the shape of Yud’s conservatism will skew towards a) conserving the pop culture he lives and breathes and b) conserving the idea that computers are (or can become) inscrutably superintelligent entities that are wholly separate entities from their users, rather than tools for humans. He’ll probably always find an audience among SaaS CEOs who are desperate to cling to their annual recurring revenue.
and if that didn’t get everyone, the food-safety AI plus MAHA will pick up the stragglers with natural bacteria cultivated in abundance
I continually notice that the assertion some particular skill is needed to properly utilize the LLM, is rarely accompanied by any meaningful description of the path to develop that skill. Maybe I’m just looking in the wrong places.