gavin leech

Hard to believe how bad this piece is; each answer contains at least one straightforward falsehood or fallacy

"Yoshua is a little different; he takes action and does things that directly focus on safety." Lots of people take action. And almost no one has ever taken such effective action as Coxon did. shallowreview.ai/overview

Hard to believe how bad this piece is; each answer contains at least one straightforward falsehood or fallacy

(I can agree, though, that it is stupid to make superhuman hackers in the first place, and stupid to run security as they have been.)

Hard to believe how bad this piece is; each answer contains at least one straightforward falsehood or fallacy

“None of the steps in this argument is credible” If Jordan is using “brain” as a metaphor for neural nets, then a weak form of the above, data parallelism, happens every day. (Hinton does believe in swarm updates - not obvious or publicly known, but highly credible) www.reddit.com/r/agi/commen...

Hard to believe how bad this piece is; each answer contains at least one straightforward falsehood or fallacy

“Hinton believes that we have figured out how the brain works” No, he thinks the opposite. analyticsindiamag.com/ai-news-upda...

Hard to believe how bad this piece is; each answer contains at least one straightforward falsehood or fallacy

The use of “young” as an ad hominem is also surprisingly miserable, for a teacher.

Hard to believe how bad this piece is; each answer contains at least one straightforward falsehood or fallacy

“Does the existential threat from AI reflect a real concern that you observe on a daily basis among young researchers?... Not among researchers.” Wrong: 51% of researchers assign >10% to human extinction or similarly bad AI outcomes. aiimpacts.org/wp-content/u...

Hard to believe how bad this piece is; each answer contains at least one straightforward falsehood or fallacy

New overview of what we're calling OpenAI's "diamond dump" of 722 papers. Includes explainers for our favourite results and also the wailing and rending of flesh. www.paradigm3.org/research/mas...

Catching up: * I started a research org, www.paradigm3.org * We have a biweekly newsletter which also inputs into our risk model p3humansonai.substack.com * You can query the dataset with Haiku www.paradigm3.org/year-in-ai * www.paradigm3.org/research/pok... * cheats.arbresearch.com

You can't buy melatonin in the UK. In the US, it's hard to avoid: it's in gas station soft drinks. How general is this kind of thing? Our new beige site looks at every case where medical regulators disagree about whether to approve a medicine. We found 101 instances www.juriscription.info/results

We did all the data work before Claude Code existed; I hope to trick one of you into doing better than us. Should take a weekend (:

You can't buy melatonin in the UK. In the US, it's hard to avoid: it's in gas station soft drinks. How general is this kind of thing? Our new beige site looks at every case where medical regulators disagree about whether to approve a medicine. We found 101 instances www.juriscription.info/results

Serious caveats: We manually inspected the outputs of the pipeline, but there are 200,000 rows here and we're not pharmacists. The 2026 models would do a lot better but we're out of patience. It's not obvious what fraction of discordances *should* resolve to universal approval

You can't buy melatonin in the UK. In the US, it's hard to avoid: it's in gas station soft drinks. How general is this kind of thing? Our new beige site looks at every case where medical regulators disagree about whether to approve a medicine. We found 101 instances www.juriscription.info/results

* If you like data gore, this is the worst data I've seen in a long career, and lives are at stake here * If you've just moved countries, you can also look up the equivalent of your favs

You can't buy melatonin in the UK. In the US, it's hard to avoid: it's in gas station soft drinks. How general is this kind of thing? Our new beige site looks at every case where medical regulators disagree about whether to approve a medicine. We found 101 instances www.juriscription.info/results

Still, over a full damn year we had a go with huge amounts of manual work on top of Claude's virtual work. We only got around to doing US, UK, and EU. Caveat aegrotus

You can't buy melatonin in the UK. In the US, it's hard to avoid: it's in gas station soft drinks. How general is this kind of thing? Our new beige site looks at every case where medical regulators disagree about whether to approve a medicine. We found 101 instances www.juriscription.info/results

Various countries have "regulatory recognition agreements", where they can reuse each other's work and get things approved quickly if it's already approved in a sensible jurisdiction. So THEN (one thought) one could use these agreements to get hundreds of meds approved elsewhere

You can't buy melatonin in the UK. In the US, it's hard to avoid: it's in gas station soft drinks. How general is this kind of thing? Our new beige site looks at every case where medical regulators disagree about whether to approve a medicine. We found 101 instances www.juriscription.info/results

A while back I noticed a bunch of "discordances" (medicines banned or unapproved in one country, but approved, even OTC, in another)

A jaundiced thread on OpenAI's (broadly quite good) New Deal policies. *Most* of these ideas have massive unflagged conflicts of interest