René Lapin

@yegordb I'll bite

Full room for our new class CS 321M: AI Measurement Science today! Why the class matters: Every field that ignored measurement science paid the price Psychometrics ran into a simple problem: - a low score could mean a bad student… or just a harder test - That forced them to build better ways to measure ability instead of just raw scores Finance learned it in 2008 - Toxic mortgages were labeled safe. The ratings looked precise, but they didn’t reflect reality - As long as the numbers looked good, the system kept going… until it unraveled Medicine is the clearest case - In the CAST trial, drugs reduced irregular heartbeats. On paper, that looked like success - In reality, more patients died. The metric improved, but the outcome got worse In each case, the metrics looked right while the outcomes were not AI is starting to look similar: - Benchmarks go up and leaderboards improve - But models shift, scores drift away from real-world use, and systems get optimized for the metric instead of the outcome We’re already using these numbers (benchmarks, metrics) to make decisions about deployment, regulation, and trust. Measurement is the foundation of science. If it’s off, everything built on it is too. We’ll be sharing more distilled insights from the class - make sure to follow along! Comment if you want the full materials (slides, textbook, etc.) (And if you commented on my last post - we're working hard to get the materials to you!)

Want to break into AI safety research? MARS 4.0 is looking for mentors & mentees for its upcoming Dec-Feb cohort, w/ an all-expenses-paid week in Cambridge 🇬🇧 Tracks in technical safety, governance, and journalism. Past participants have published at EMNLP, NeurIPS, and ICLR. Deadline is Oct 1st. 🔗👇

This year, a Georgia court sentenced a man who abused 107 dogs to 475 years in prison. Also, this year, US factory farms confined over 70 million pigs in equally abusive conditions. None were prosecuted. Imagine if we enforced our anti-cruelty laws against all animal cruelty.

@robertwiblin though an american i've _never_ heard the 'american' version.

That nonplussed means the exact opposite in the US vs the UK is really quite confusing.

An explicitly illiberal project built on the delusion of achieving permanent political and ideological dominance. Politicize everything. Deploy lawfare against your political rivals (and anyone standing too close to them). Purge dissent. Expand executive power, even at the expense of constitutional fidelity. And most importantly, assume (against all reason) that once seized, power will remain safely in your hands — and that if it somehow slips away, your enemies will suddenly rediscover restraint. It’s an astoundingly shortsighted and tragically common form of self-deception: the belief that you can normalize abuses without eventually being consumed by them. *** The rhetoric is almost comically bellicose (the ⚔️ emoji radiates *scrawny theater kid who always dreamed of making the varsity football squad* energy), but bravado isn’t strength. And delusions of grandeur are not a strategy. The Trump administration isn’t just embracing this approach — it’s already demonstrating how incoherent and self-defeating it is in practice. These aren’t “new tactics.” It’s a clumsy retread of old, doomed ideas — the kind that leave you weaker than when you started. “Get ready.” Diabolically cringe. Classical liberalism isn’t magic. It doesn’t defend or renew itself. Every generation must relearn why it matters — and fight to safeguard it against threats foreign and domestic, and perhaps especially domestic. Whatever its defects, I prefer classical liberalism to Rufo’s prescription. He is part of the problem. Been true for some time.

@18dixiedean78 @mattyglesias

@mattyglesias The Rent Control of international economy!

@What46HasDone @KelseyTuoc "there's nobody in this world more useless than a blue maga biden anon. who gives a flying fuck about you or what you have to say" we can do this all day.

The first of my demands for a takeover of the Democratic party was a blackballing of every single person who concealed that Biden was in cognitive decline and obviously not capable of serving for four more years. A bunch of people questioned this demand.

@itaisher Sadly I'm going to have to unfollow Itai. The tweet never happened.

PEPFAR has enjoyed bipartisan support for decades because cheaply saving the lives of millions of children is good and the right thing to do. If Congress decided to end it, I'd be devastated. But that's not even what's happening. Congress authorized it! Our democratic government decided to spend money on this! The state department has ordered a stop-work on funding that Congress did authorize, because 'stop all foreign aid' sounded good to someone. It's not clear that anybody wanted to let tens of thousands of babies die of HIV/AIDS. My best guess is that they just, you know, didn't really know what all of the programs they were ordering halted did. I'm old enough (like...two years old...) to remember when Republicans considered themselves the party of saving babies' lives. I defended the pro-life movement against charges that they really just hate women, saying that I've been honored to know a lot of pro-lifers who sincerely want to save babies. But if you can't bring yourself to take a stand on this - that the state department should not abruptly halt a program half a million sick children depend on, which Congress authorized and paid for, with no justification, in a way that will make the problem much much worse when work on it starts again - then you aren't pro-life, and you never were.

@lethalrejection wait so she herself is the freaky hacker chick? that’s the solution?

@michaelj505 thank you for doing the lord's work here

Overall, this has been a very interesting but also challenging experience for me, and I would like to thank all colleagues and friends who supported me.

Is a study robust when its findings depend on the inclusion of a single observation? In a new report, @Ben_Guinaudeau and I discuss this question based on the replication of an @apsrjournal article. A thread on our main findings.

Just had a brief but very illuminating conversation with President Biden, whom I support fully. He told me as a kid in Scranton he used to… here’s the deal: you take an old tire and fill it with soup - listen. Ten, twenty people in one car. You better believe it. Pepperidge farm

@mattyglesias * employed by harvard as a social scientist

@mattyglesias how can you be at harvard and bungle the mediation/moderation distinction *sigh*

This paper by a political scientist who has since moved on to focus primarily on administrative work is quietly devastating to the delusions that have undermined the last decade of progressive politics. papers.ssrn.com/sol3/papers.cf…

@RachelBitecofer @mattyglesias (he's making a humorous remark -- note the faux grandiosity of "Battle of Interpretations")

@mattyglesias That's what you're worried about?! Trust me, there isn't a problem you have right now that won't seem quaint if Trump becomes dictator.

If Biden loses (which seems more likely than not at this point) I’m probably going to lose the subsequent Battle of Interpretations and people will decide the idea of nominating an electoral moderate is discredited. Long winter coming.

@AssLatam I enjoy this account … but I’m pretty sure you just got absolutely played by a meme-maker :/

Keiko Fujimori, former Peruvian presidential candidate, shows the results of her recent piano practice by playing "Imperial March" from Star Wars.

About to do some machine learning. Wish me luck.

2day Barbara had a pork tenderloin in Bondurant w Fred yes the Fred who hit a deer near dyersville that I assumed dead

@analisapackham Mexico City is the correct answer!

I have been sufficiently bullied into taking a few days to go somewhere in March just for fun. Please help me decide. I don't really care about outdoorsy stuff but would like to go to a city with good food/drinks that's walkable + safe at night. Any and all insights appreciated!

@Noahpinion definitely had a conversation about this in mexico as well. so, as many people in the comments are pointing out, not only usa. an old psych professor's comment about this was that loud music causes people to lean in closer and get flirtier. makes little economic sense, tho

Why do American bars all play really loud music, even if everyone is just sitting at tables trying to chat?