Andrew West

ah damn, RIP Margaret Hamilton. she took safety-critical software engineering seriously, and got people to the moon and back. news.mit.edu/2026/margare...

Some early first-hand accounts of the experience of encountering a narrow superhuman intelligence as mathematicians grapple with the hundreds of big AI proofs released by OpenAI. Problems solved in inhuman ways that make us wonder what it means to actually know things... scottaaronson.blog?p=10169

Very important to watch yesterday’s big speech so you know what’s going to happen to the UK’s productivity figures in the next year. Nice of OpenAI not to clash with whatever Andy Burnham was saying too.

good to hear. I think it’s clear that if you know how to learn and want to learn, AI will not impede you and will most likely help. what I don’t know is what proportion of people are like that. people talk about college kids like they have to be forced to learn, or tricked into it. isn’t that weird?

I don't understand the need to say dickwad CEOs can't make good stuff. This is the world we live in. A lot of my favorite things in the arts and sciences come from absolute bastards. The real argument should be about how to channel dickwaddery into good social outcomes, not whether the stuff works.

So many people with aviation psychosis claiming that the Wrights’ invention is going to reshape global geography when it can’t even go more than 25 yards

Muting the word ‘billionaire’ is very effective, I recommend

The Guardian’s data centre reporting is needlessly hysterical (FREE TO READ!) takes.jamesomalley.co.uk/p/the-guardi...

It’s just predicting the next tokens after “the paper I would submit to win a millennium prize is” openai.com/index/navier...

Hard to see what BBC breaking news notifications are achieving at this point

Any MP who signs on to Burgon's motion is not acting within the national interest. We need more data centres – and it comes down to a question of security: takes.jamesomalley.co.uk/p/the-case-f...

The Imagination Library is a truly wonderful thing, a beautiful idea that meant so much to so many people at home and all over the world

Gonna briefly do an AI take: it used to be possible to have interesting discussions about AI here, and it's got progressively less so. You can't just think AI is bad, or overhyped, you've also got to subscribe to a very particular worst-case scenario about *how* it's bad and how bad it is.

Last August was dry, and I had to stay up late one night monitoring my parents’ neighbour burning his scrub in an impromptu bonfire. Nice line of trees between me and him. Little jolt of adrenaline every time a spark went their way. Some people are just not that bright and don’t watch the news.

So that really triggered the mad libertarians huh

Been testing this 'sign-language-to-text' (SL2T) for almost a year & it's so awesome to see this out! Looked at this almost a decade ago & knew we'd get the tech & ML models there eventually (but super challenging). Talking w/ some Deaf folks, an equivalent of 'voice-to-text' was huge for them

Just walked to a more westerly window in order to be closer to the sun

BIG NEW piece by me. A few years ago, if you asked me to summarize my entire worldview in one article, I would’ve said it was too broad and too difficult. But this year, I decided to do it.

I get that the sirens’ song is inherently not composable. But Christopher Nolan has infinite money and any artist in the world would take his phone call. Maybe try *something*. Hell, get AI to invent a musical psychology and compose something that induces a compulsion oh no wait

There's lots of stuff LLMs can't do, but if you want to disbelieve all the recent math, you have to believe a bunch of Fields Medalists with long track records are just straightforwardly lying about the thing they care most about. Here's a recent blog by Gowers: gowers.wordpress.com/2026/05/08/a...

Tucked away in this article is an appeal to the AI skeptics to PLEASE stop writing off stories like this OpenAI accidental exploit of Hugging Face as a dishonest marketing trick Frontier models can find and exploit vulnerabilities now, it helps nobody to pretend that they can't!

the problem with AI is that it works, not that it doesn’t

Tech commentary often has to be ‘what if now, but forever’, and that doesn’t work well when change is so rapid. An easy flipside is discussing power dynamics rather than the tech itself, and that’s very effective at riling people up, but the world still turns.

"In mid-2026, autonomy is not just some hypothetical or highly experimental technology. Self-driving cars have already been invented, and are a thing that have existed in the world for some time."

Since we can exercise discretion about breaking news alerts for overnight football matches, maybe fewer murder alerts at Christmas?

I honestly thought I would never see the day workspaceupdates.googleblog.com/2026/06/work...

Quite fun watching the normie press discover ramageddon now that Apple has put its prices up. Thankfully there’s 6 months of deep dissection in tech podcasts that will help the coverage be sensible and informed.

It is weird that there is still a substantial set of people who believe "AI is mostly hype" at this stage: Five Eyes is warning about AI, exponential revenue & token use at the AI labs, unit distance/Erdos proofs, and so on... There are many real issues with AI, not being real is not one of them.

“I am appalled at the man blasting music over this very serious occasion. Anyway now back to gleefully reporting on successfully hounding another Prime Minister till they cried.”

I love the smell of a moral panic in the morning. (A surefire sign you're hitting peak moral panic and heading into the backlash: thing X gets called "the new smoking". Nothing is *ever* the new smoking. Nothing ever will be again.)

Always good to have your response to a consultation immediately ready to go. Suggests you've definitely engaged with the results of the consultation and that the entire thing wasn't a waste of time.

I seem to be a bit of a contrarian on this and yet it seems to distort a lot of discourse: at the current stage, the model scale and "the weights" are only part of the capability story. The scaffold/harness, the memory systems, the files, and the pilot themself also cause very divergent outcomes 1/6

Weird how GDPR doesn’t apply during elections

Trying to think up examples of media that has tipped over into myth. Where people fully act like it really happened, despite sorta knowing it’s fiction. Lord of the Flies is a big one, and Romeo & Juliet. But also: Love Actually

I'm a supporter of the Lords and in particular of the way that it controls its own timetable. But this is a constitutional outrage and it cannot stand humanists.uk/2026/04/24/a...

I would love to know why effective altruism of all things triggered the mad conspiracy gene in so many people