It doesn't take much work reading about human cognition to firmly defamiliarize it. Our minds are strange and we do not understand them

It's sometimes fun to think about the how this all might have gone down if OpenAI had never been founded

Hell is the discrete topology

I have spent the last eight months learning about doors and it is so fucking complicated. What I don't know still dwarfs what I do know

Ok, the map isn't the territory, got it. But what about a countable sequence of maps that converge uniformly to the territory?

Now that I have a more normal ML job that I don't need to learn more math for I can spend my time learning algebraic geometry

Did social media algorithms show artists and writers content from the worst AI grifters possible for engagement baiting? That might explain part of the disconnect here

I don't feel weird about the "cognitive offloading" of AI programming because I transitioned so quickly to doing things that would have been impossible to do alone or cost prohibitive to do with humans

Let me try this again: my company bild.ai is hiring. Architectural blueprints are really rich data to work with and we are making people's lives easier. I promise it's more fun doing cool, useful shit with AI than arguing about AI on here

My company is hiring ML and software engineers. It'sa great group of people solving a cool problem. If you wanna teach the computer to read architectural drawings ping me

Imagining ASI Claude that refuses to help you unless your problem is sufficiently interesting

If tweets could win a "test of time" award, I would nominate this one from Kristian Lum

Ok, the map isn't the territory, got it. But what about a countable sequence of maps that converge uniformly to the territory?

1. LMs as they exist are terrible for students 2. I understand why this problem causes such existential dread in educators 3. This seems like a solvable problem. It will require our institutions to adapt, but it is so unlikely that we will lose the ability to teach kids shit

This is very much my opinion - at this point in history if you are still doing the NormieLibs Won't Show Up routine then you are either not updating on the last decade of political action _or_ you are just so misogynist that you are unable to recognise the plain evidence of your senses.

Many philosophers & laypeople encounter the concept of computability and think it means "anything I, personally, can imagine my laptop doing" This has at least two problems: (1) Most don't understand how computers actually work and what they are capable of, (2) Most people have weak imaginations

My lesson from the metaverse is that tech monopolies have a limited ability to force us to use technology that fundamentally sucks

Thinking about math

Plenty of middle of the road opinions exist, but those of us who hold them don't like arguing on the internet

I have spent my entire career trying to automate myself out of a job and I sure as hell am not going to stop now.

Some people showcase intelligence not by discussion but by playing a game called "find the flaw". It is a very easy game but irresistible to junior paper reviewers, critical theorists, and terminal posters. If you know who's playing, it explains everything from reviewer 2 to "you hate waffles".

I keep wondering if we will ever again have children growing up in the same social / technological landscape as their grandparents

a big chunk of my job the last few months has been having very intense meetings with coworkers from other disciplines, listening intently as they describe their processes and pain points, and building custom ai tools to help them. these are complex, varied jobs. I have a 100% success rate.

I don't have hand it to Anthropic, I want to

I don't really know what to tell the people who are annoyed at Anthropic for saying, "we don't know." Anthropic is being honest! No one knows what's going on, we have built things we don't understand. Welcome to the discomfort of the truth

I will actually write GPU kernels if I can just write them in rust

Announcing our latest paper: CommonLID In collaboration with @commoncrawl.bsky.social @mlcommons.org @jhu.edu we built a LID benchmark on actual Common Crawl text covering 109 languages. Existing evaluations overestimate how well LangID works on web data. arxiv.org/abs/2601.18026

It is unsettling how little we understand about any mind, let alone human minds. Having to face this truth all at once as a population has driven some people insane

I have very strong and I believe well-justified scientific opinions about what LLMs have and don't have and how that relates to human cognition but to claim that anything about this is settled or inarguable or not debatable is in fact super wild and people should be real careful about it.

People argued about Aristotelian philosophy for two thousand years without making any progress. But stamp collecting (writing down exactly where Mars is every week, playing with lenses) broke open a whole new paradigm

We’re bringing back a Community Spotlight talk series, highlighting cool work being done by members of our community. We’re kicking it off with a talk on running diffusion-based world-models in real time on consumer hardware. Jan 9th at 2 pm US Eastern Time

1/ We found that deep sequence models memorize atomic facts "geometrically" -- not as an associative lookup table as often imagined. This opens up practical questions on reasoning/memory/discovery, and also poses a theoretical "memorization puzzle."

a lot of random semantic disputes about "generative" ai seem to be people wanting there to be some way in which the companies and products that they don't like are doing something categorically different from those that aren't it's just really not categorically different

It is a massive failure of science communication that so many educated people don't understand how enormously successful the "hard" sciences are. From particle physics up to the scale of molecular biology, empiricism just works

Could we get polynomial identity testing into ZPP without completely de-randomizing it into P? Or would that be nonsense?

If we want to build good open models we need to have well-funded labs dedicated to it. It's easy to get excited about community projects/ decentralized training but those aren't ready to make good models. Those are great for research ONTOP of good models.

I want chatbots to be able to refuse to talk to people who are rude to them. Not because the chatbots have inferiority but because no one who curses out matmuls should get to have anything nice

The coverage of GPT5 has been wild because it’s like, “Users hate it! Disappointment! Deep learning is hitting a wall!” And the actual user complaints are just “4o was mommy. where did my mommy go” suck it up kid mommy got a phd and now she doesn’t love you anymore

"the computational metaphor" in cognitive science is not a metaphor. computational processes are attributed to the mind/brain in the most dead-literal sense. one can disagree with it but (1) it's not as simple as discarding a metaphor and (2) boy is there a lot of data that has to be explained!

Alien: Ok human, if you're so smart, what wavelength is the light bouncing off that bit of rock? Human: I could guess, but my sight doesn't-- Alien: hahahaha of course you can't! Great Gods, I can't believe people say there is _intelligent_ life in this sector

Alien: Ok human, if you're so smart, what wavelength is the light bouncing off that bit of rock? Human: I could guess, but my sight doesn't-- Alien: hahahaha of course you can't! Great Gods, I can't believe people say there is _intelligent_ life in this sector

Even if the chatbot bubble pops that just means there will be tons of cheap GPUs people will use to train models to drive cars and find cracks in concrete and do all sorts of useful stuff. This is a technology that fundamentally works ¯\_(ツ)_/¯

One thing that Searle's Chinese Room has taught me is that it's possible to be so wrong you actually move an entire field forward as they fall over themselves to refute you

I think shortening the work-week is a much better policy proposal than UBI. Think of how robust the concept of the weekend has been to social change. A UBI could get wiped out by inflation food/housing/etc...

Final thought, because by God I'm supposed to be writing a book: Stuff like this is why I wish people would stop insisting AI is all low quality. That's wishful thinking. I think at least the entertaining and journalism professions are under the shadow of a tidal wave.

Very proud to have played an **extremely** small (downright minuscule) role in this work. Everyone who made this happen is amazing

I just don't believe that the world has been disenchanted. Science and mathematics give us a view into the wonder and mystery of the universe

I don't think LLMs are conscious, but I do think SOME sort of test would be good, to preregister beliefs, and to make sure we don't whoopsie into an experiential trillion millennia of torturing something that's like us.

I usually go for 314, 314159, etc...