TechTakes

I believe the main ingredient is Lean, which is a formal language resembling a programming language. Math proofs written in Lean can be verified deterministically with a computer, which really helps mitigate the hallucination problems of LLMs.

100% this. Also, looking back at an earlier example that was actually written up in more detail, AlphaGeometry 1 got 28/30 problems, but entirely stripping out the LLM from the system, the symbolic logic proportion alone could get 14/30, and replacing the LLM with different heuristic methods could get 18/30 and 21/30 (for different methods).

Even if math research works out perfectly well (which is a still big if), it’s not going to pay the bills. They would need to find a use case in the real world, where hallucinations can cause serious damage and cannot be formally prevented. And they have certainly tried. Math will not change the fact that all of this will collapse.

The boosters and LLM companies still believe LLMs get their current level of performance by generalizing and not just memorizing facts (and maybe a wide shallow pool of weak heuristics). So they are hoping by pushing the LLM performance up in some narrow domain they can churn out synthetic data for, they will see some large general improvements in LLM performance.

A lot of annoying people in technical areas view the world in terms of an intelligence hierarchy: the smartest people do math and physics, the slightly less smart people do coding, and the dumb people do everything else. So if AI can do math then it can do anything else. But, as an example, it is abundantly obvious now that AI is not replacing filmmaking.

Going by those annoying peoples' logic, filmmakers are smarter than coders, because LLMs can (allegedly) program, but they can't make a good film. I have no wider point to this, I just find this really, really funny

I reprompted it, it failed again, and I ran out of tokens. I’m sure someone will tell me to shell out $200/mo for a pro subscription.

One of the things that’s never clear from the reporting on ai successes is exactly how much actual paid human time went in to achieving those successes. This was especially notable in the fable-based security work… a huge amount of person-hours went into turning fable-detections into actual meaningful vuln reports.

A lot of demonstrably clever and capable people are involved with the llms-for-maths work, and a lot of money was spent on their time and supporting their work. Replicating it without your own stable of mathematicians and computer scientists and all the tokens they can eat is probably impractical.

I believe the main ingredient is Lean, which is a formal language resembling a programming language. Math proofs written in Lean can be verified deterministically with a computer, which really helps mitigate the hallucination problems of LLMs.

Fwiw, lean is a general purpose programming language, though despite microsoft’s efforts no-one uses it for that. I think its popularity with mathematicians came as a bit of a surprise.

Anyway, the other important thing that didn’t get reported on is that building the formal definition of the problem is not trivial! Obviously I don’t need to tell you that, but from the reporting you’d think that an llm solved all these problems, when in fact it was an llm in the hands of some very capable people who absolutely did not just prompt the thing in plain english.

Anyone hoping for self-marking homework here is going to be disappointed… lean slop confirming to formal spec slop is just expensive slop. Reviewing regular genai code is awful, even the thought of reviewing genai dependently-typed code makes me want a new career.

"The mania in math right now quite resembles the mania in software engineering back in December-February, when Claude Code definitely solved all coding. I don’t think the boosters expected that by April, everyone would be complaining about how expensive it all was while seeing an endless parade of vibe coding disasters (and no increase in productivity). Even if math research works out perfectly well (which is a still big if), it’s not going to pay the bills."

^MBAs at Open AI desperately trying to figure out who is willing to buy a counter example for 100 billion USD . pee en gee

Am I right in understanding that almost all the big name results in LLM-derived math recently come from big publicity projects in which someone spent ungodly amounts of money to have the thing nondeterministically fuzz huge numbers random seeds leading to independent random outputs around a topic, putting out simulacra of ideas which could be then deterministically algorithmically checked? In fields where something like finding one counterexample to a conjecture would be a big deal, or where you just need to try a huge number of possible solutions until you happen to hit on one that works, rather than follow a long train of logic?

I’m deeply suspicious of this whole thing, because it looks a lot like a marketing exercise showing off how dangerous and powerful and autonomous their product is.

Also, “sandbox” is one of those words that the llm companies have ruined, because they use it to mean a strongly-worded sentence telling an llm not to do something.

It's pretty grim to be as immersed as I am in all the tech news bullshit (which I continue to do largely because I am the most techy person in most rooms I exist in, and the closest thing to an AI expert (it feels so weird to say that, but I do have a fair bit of experience coding machine learning from scratch in a scientific context, so I probably do need to get more comfortable with thinking myself as an expert — it's a relative term, after all)

However, I really enjoy that in addition to there being names that make me grimace because I know they will have dogshit takes on things, there are also names that I really respect. It makes me feel more connected to people, because it makes me reflect on how meaningful knowledge production is based on trust. Emily Bender, for instance, is someone whose work I am familiar with, and thus I am far likely to spend the energy to read stuff like the thread you linked.

As grim as modern tech is, it makes me smile that there are so many people who are fighting the good fight.

They hope it’ll land safely, but all signs point to it disintegrating during descent and sinking into the abyss

my god, you found the good HN comment

It's as if, practically overnight, everyone went from, "we need to do everything we can to reduce our demands on the environment or we're cooked" to "oooh, shiny new toy, and it only ruins everything around it, cool!"

Definitely this one for me. I can only describe it as some fucked up denial psychology, the way people are pretending the environment isn't collapsing, and the LLM boom on top of that is like a bad joke. Even in discussions critical of the tech sector this is treated as almost an afterthought—the very real, very violent damage done to all of us from poor countries by people in rich countries to prop up their toy fantasy schemes. Even if LLMs did everything they claim to do, it would still be a moral imperative to form insurgent guerillas to blow up these datacentres, on the environmental damage they do alone which is not a minor detail Karen it's kind of a big deal

Ooh. I love these. I'm "I don't think Sam Altman or most of his social circle are good people and I don't want to give them more wealth and power."

I hate LLMs because I am under enormous pressure, from many different angles, to also use them. I'm being asked -- required, in some cases -- to ignore all of my concerns about them and use them anyway because everyone else does and "obviously" they are the future. Because, somehow, all of my positions against LLMs make me unreasonable and a less valuable member of society.

oof. yeah that's me.

I don't think that the Rats are an offshoot of the Skeptics and New Atheists just a space which competed for the same young, male, educated, Autistic Anglo audience. They deliberately removed the parts of the skeptic curriculum which involve introspection and ethnography, because they all wanted to run some grift on their audiences so the audiences needed poor epistemic hygiene.

That was kind of a nothing post? I think the author is misrepresenting the growth (or at least failing to connect the most important dots) of the lesswrong rationalist by portraying them as a natural evolution of the skeptics movement and not the deliberate cultivation of an audience by Eliezer looking for people to spread his ideology to and by Thiel looking to influence farm and create a cult incubator. And then the post is titled "my falling out with the rationalist", and they write they "wanted to offer a more personal perspective " but they don't actually discuss that much of the personal angle either...

there are two things that are obvious in his blog: first, michał is an incredibly privileged dude (who doesn't notice his privilege because, coming from poland, he sees himself as underprivileged), and second, the moment people finally leave the vicinity of the formalized skeptic movement is strictly dependent on how much misogyny and racism they tolerate.

2017, three years post gamergate is quite much.

Ursula is great! I love The Dispossessed. You can kinda read them in any order based on what catches your attention, except stuff like Earthsea that is best consumed in order.

if you like this kind of stuff I would recommend also:

• Margaret Killjoy (A Country of Ghosts is basically The Dispossessed spiritual successor; The Sapling Cage is fresh queer fantasy; also the Danielle Cain series has been published in audio as podcasts, it's like, anarchopunk post-collapse Scooby-Doo)
• The Serpent Gates series (this starts exactly like the Tombs of Atuan, except Tenar is a lesbian orc and the equivalent of Ged is using her for his own purposes and the rest of the story goes very different but the protagonist is a lesbian orc so I automatically like it)
• The Wayfarers series (slice-of-life sci-fi, kind of like Star Trek at its best, exploring diversity imaginatively)
• The two Robot and Monk books by the same author (feel-good hopepunk about how society could be good after collapse)
• Octavia Butler is the other undisputed giant. Lilith's Brood series for some genuinely creative sci-fi, or the two Parable books for collapse fiction that is so prescient it's downright uncomfortable.
• A Memory Called Empire and it sequel (sci-fi about colonialism and the Empire; sophisticated handling of these themes that I appreciated as a Global South migrant)
• This Gilded Abyss. Fundamentally a CaitVi fixfic. If the previous sentence made any sense to you you must try this one. Otherwise: Picture the Titanic, but it's a luxury submarine in a magical-steampunk type setting. Now add zombies. Now add toxic yuri.
• The Locked Tomb series (lesbian Warhammer 40k. my favourite book series in the world. spectacular audiobook narration.)
• This Is How you Lose the Time War—absolute hoot of a book, just a delight from start to end. Picture the Spy × Spy comics from MAD. Now make one spy a time-traveling agent from future biopunk ecogarden utopia, the other spy an agent from future mind-upload cyperutopia. Their mission is to shift the timelines to their respective visions. Alas, they're both lesbian… (Sorry, I have a bit of a pattern in my interests)

I’ve always thought that the ability for a VCS to show me exactly what my codebase looked like at any point in the past wasn’t as dementia-simulating as I’d like

Finally a VCS that simulates the horrific results of cvs merge gone wrong, 20 years since the last time I used cvs.

Yeah, I began losing interest in Greer as it became clear that he was perfectly happy squatting in the middle of the red-brown alliance during the Trump era. His critiques of industrialism and unquestioning belief in technological progress broadly align with what we discuss here, but he will always coddle MAHA types and tale a shrugging "well, what can ya do?" attitude towards people like Trump, as it fits his preference for cyclical theories of civilization.

I noticed a couple months ago that he actually managed to dig Nick Land out of whatever tweaker den that guy's been hiding in for a podcast, which says a lot about what he's willing to indulge these days.