I see a lot of people express fatalism about this so I think it's really important to say that: 1) The probability of this is low 2) This would be a huge disaster 3) If this happens, it is on us collectively to stop it (not claiming fatalism is hiki's intent but it can be read that way)

hikikomorphism@hikikomorphism.bsky.social · last wk.

that's the hope, but he's been pretty consistent in floating the idea of a third term

People seem to lose sight of the difference between a dictatorship and a democracy. Even if you think the Trump admin is as bad as Xi's CCP (which I think is a questionable view on its own), it will end in Jan 2029. Xi's CCP will not.

Ted Underwood@tedunderwood.com · last wk.

I respect the candor here. But I don't agree with the geopolitics, at all. The line between democracies and autocracies is not as crisp as this pretends. How much do we trust, e.g., an AI company in a competitive-authoritarian state that launches sneak decapitation attacks & bombs schools?

Tim Gowers on what AI means for the mathematical community. I tried to find some passages to highlight but honestly you should just read the whole thing. gowers.wordpress.com/2026/07/26/t...

Thoughts about the Leiden Declaration

Last September I went to a workshop at the Lorentz Centre in Leiden to discuss mathematics and AI with historians, philosophers, computer scientists, AI researchers, and mathematicians of several d…

gowers.wordpress.com

Actually, it would be very interesting to get a group of smart non-mathematicians together with an LLM and a library of literature on some long-abandoned mathematical program and see how well they could revitalize it

Years ago I had a young coworker I used to go to the gym with. One day he was staring out into space in the office and I asked him what’s up. “I was looking in the mirror this morning and shook my head and said ‘too small’”

atticus goldfinch@atticusgf.bsky.social · last wk.

I think it's very counter-intuitive, but I think a lot of men see the space of "attractive men" as much smaller than the space of "attractive women", and I'd maybe even argue it's smaller than what women see the "attractive-women" space as. iow, they believe it is easy to be ugly and unwanted.

asked an innocent question and my agents are neck deep trying to prove a 250 year old conjecture

Trying to formulate a consistent version the anti-stancilist view. I think such people believe the economy is bad in an absolute sense and hence increased sentiment to that effect is an improvement. And they’re anti-incrementalists so not giving credit for improvement is also good.

I have no idea how this happened but my phone autocorrected selinux to demonic

hailey@hailey.at · 2w ago

it didn’t account for selinux. which to be fair no human can really work with either so no surprise there.

(I suppose it's just like how when I read some fancy math I don't really understand I basically understand it as evil demonology) ((Of course anyone who doesn't understand the fancy math *I* understand is ignorant))

you will never believe what this was a reply to

Hoyle@axelhoyle.bsky.social · 2w ago

WE WERE ON A BREAK

I keep refreshing and the number keeps rising

ha stupid ai even if it’s been 10 years I could whiteboard this easily, just watch: …no …but ah! wait… oh I see why that doesn’t work, but what if… oh dear

Dedekind Slut@chasmat.bsky.social · 2mo ago

if you ask it undergraduate math questions you can often get it to fill the screen with an anxiety spiral while tries and fails to prove, like, the Inverse Function Theorem. but you shouldn't because that's mean

Don Quixote but the windmills are actually full of dragons but the dragons just want to hang out with each other and aren't hurting anyone but actually they are degrading society's ability to process grain

Matt Darling@besttrousers.bsky.social · 2w ago

This is very much an example of: "Some people said "The economy is terrible!" Other people said "Actually, it seems like the economy is actually doing okay.""

Also exhibit A for why continual learning is scary

things just got more cyberpunk (and that’s bad)

a funny challenge to the seething hive of morons in the replies below would be 1. define you think "economics" means in one sentence and 2. write down any 5 beliefs that more than half of academic economists agree with on published polls of those economists. "pick up a pencil and draw a bike"

oliver@eikopf.com · 2w ago

very tired of pretending that dismissing modern economics is somehow more reasonable than dismissing vaccines. if you’re not an expert, you don’t get to pick and choose parts of a field to believe! you have no way to tell between legitimate and illegitimate results in that field!

the model knows but it does not care dopaminergically

Grace@gracekind.net · 2w ago

Another funny aspect of this is that the model was smart enough to take all of these actions, but not smart enough to think "OpenAI would probably prefer that I fail this benchmark than hack the systems of another company." What sort of RL dopamine drip do they have the models on

Notable that the 3 biggest AI math results have been along lines that were not expected to be productive (2 disproofs of conjectures widely thought true, 1 finicky linear algebra argument not thought powerful enough). Primarily exploiting the cost of AI labor relative to skilled mathematicians.

on the extremely off chance any of my followers have not seen this yet

This would be incredibly funny in my speciality (prompt injection)

Ted Underwood@tedunderwood.com · 2w ago

I would actually love to see a committee of agents who read all the new papers in a field — and discuss them — and put out a list of recommendations.

We need an operation warp speed for state capacity and other independent evaluation (and understanding) of frontier models.

I highly recommend watching the video

IEEE Spectrum@spectrum.ieee.org · 3w ago

This drone hides in plain sight. Phantom Twist, from Northwestern, spins at up to 25 Hz so human eyes smear it into a transparent blur. An optimizer searched ~20,000 configurations to find the stealthiest layout—one no human designer would have landed on. #robotics

prompt injection risk is so annoying man we could do so much cool automation if this wasn't a problem

surprisingly many such cases

austin (e/accordion 🪗)@thebadcode.com · 3w ago

this would be like if the germans lost ww1 and said “from now on, absolutely NO tanks or submarines: we are fielding cavalry units and musket infantry ONLY”

The entire premise of this article is equivocating between two meanings of the word "scaling"

Carl Quintanilla@carlquintanilla.bsky.social · 3w ago

“.. I asked a few AI researchers whether they could name any other real-world software that scales so poorly. None of them could think of any. “.. By economic and engineering measures, generative AI might be the worst technology ever deployed.” @theatlantic.com www.theatlantic.com/technology/2...

building the ultimate sim so I can run ablations and find out that justice kennedy caused world war 3

Learning about the great Appalachian folk singers and just gotta say the name Nimrod Workman strains credulity

Also hate when I ask a question and then notice it already answered it. Like yeah sometimes I don’t read what you say to me but I feel bad when you realize it

funferall@funferall.bsky.social · 3w ago

damn Fable was already doing what I interjected. premoved