o4-mini-high is the first AI to pass my personal secret benchmark for hallucinations and complex reasoning, so I guess now I can tell you all what that benchmark is. It's simple: I post a complex midgame chessboard and 'mate in one'. The chessboard does not have a mate in one. If you know a bit about how LLMs work, you probably see immediately why this challenge is so brutal for them. They're trained on tons of chess puzzles. Every single one of those chess puzzles, if labelled "mate in one", has a mate in one.

The Miami Herald published a great article about one man who applied for refugee status, received it, and was seized at the airport and sent to CECOT for no reason: I posted it and everyone went "yeah that seems bad". Garcia, though, we can argue over.miamiherald.com/news/local/imm… This is the toxoplasma of rage phenomenon, where the most controversial case gets debated far more than the clearcut ones where everyone just goes "yeah okay, that's fucked up" slatestarcodex.com/2014/12/17/the…

The 20million number is an article of faith on the right, but it's completely made up. I've been trying for a long time to track down how they came up with it. Even if you assume that literally every border encounter was an illegal alien allowed in, you get half that number.

https://twitter.com/JDVance/status/1912320489261027374

This is important because they'll explain why they have to throw innocent asylum seekers in El Salvadoran prisons with 'well we have to get 20million of them out' and they don't, because half of them are imaginary people who only exist in their head.

We just finished dividing 5yo's room into two to make space for the (planned, not yet in progress) new baby. He's been enthusiastic about the whole process, and I promised him that while the room would be smaller we would make it his absolutely perfect awesome room. Two walls are painted orange and two blue, his choice of colors. Today we put up dinosaur wallpaper and I got a bunch of glow in the dark stars for his ceiling. To be specific, I got 530 of them. I put up a rocket, the Moon, three saturns, an astronaut, some beautiful stars.

8 year old just asked at the dinner table "what is a liberal" and now three adults are vehemently speaking over each other. "people on the left in the United States tend to favor more government intervention in markets-" "that's NOT liberal don't you tell her that's liberal" "kiddo, are you aware that in the past women didn't have as many rights as men? at the time, the people who wanted to change that were liberals" "that does NOT mean -" "today, BOTH liberals and conservatives are in favor of women getting the vote -"

Today in Vox, inspired by Zeynep's NYT editorial earlier this week, I wrote about why we badly need a Covid reckoning and are never, ever going to get one: Covid touched the lives of every single American. Decisions that were made by public health officials dramatically transformed our lives. They were hard calls, and no one was going to get them perfect. But worse than the mistakes is the lack of reflection on what went wrong.

I haven't had the chance to crack open my copy of Abundance yet so I've been avoiding wading into the discourse. But a thing I believe fervently is that the natural human desire for our lives, our neighborhoods, our communities and our world to be good, combined with the absurd material prosperity that we enjoy in the world today and the even more absurd material prosperity that our work can ensure is available to our children, ought to be sufficient for high and rising standards of living. We are rich enough to have spacious beautiful homes in shaded, safe neighborhoods with good schools and good jobs. We are rich enough for that to be in reach for every American who puts in the work to pursue it, and for it to be available to a growing share of the globe as well.

And we don't have it. And part of why we don't have it is because we have made it incredibly difficult and frustrating and prohibitive to build it. A couple years ago, I and some other families came together to found a microschool for our kids, because they weren't well-served by the local public schools and the local private schools were staggeringly expensive and still not that incredible for the money. We run it out of the basement unit of my neighbor's home. The kids love it. They thrive there. We offer scholarships to make it available to families who could never afford it; we use an absurd amount of parent volunteer labor to make it possible and affordable.

Doing the administrative work to make this possible was basically a 20 hour a week job for me for the last two years, with huge contributions from a close friend who is a professional accountant and another close friend who likes doing taxes. There aren't a lot of rules on the books about a microschool like ours, so we had back and forths with the Oakland department of Child Care Licensing about whether a school needs to be a licensed daycare or not which took months to resolve and involved my contacting constituent services for help. California law had thankfully created a carveout for home care services in zoning laws, or we wouldn't hav ebeen able to do this at all. We've been trying to become a nonprofit for more than a year now, so that our ongoing large contributions to fund scholarships for other kids at our school can be legally eligible as donations. California requires that by the end of this year an employer offer their employees retirement accounts even if there's only 1 employee - the only state to make the requirement kick in even for tiny businesses. Getting liability insurance was a three month odyssey.

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. Donald Trump is crushing our economy with insane tariff policies that kill investment. He's dismantling crucial services. Republicans are making housing and employment discrimination against trans people legal. He's burning our alliances. Why go after Biden's people?

What would some good unifying demands be for a hostile takeover of the Democratic party by centrists/moderates? I have some ideas....

https://twitter.com/mattyglesias/status/1892331769900257772

Every single person in the Biden administration who concealed that the President was unable to discharge his duties should be expelled from the party. (I know there's not really an expulsion mechanism; figure one out). They should never work in Democratic politics again.

PEPFAR is one of the most popular, bipartisan US foreign aid programs. The State Department says it has saved 25million lives, but there isn't much public, independent verification. Last week I invited some friends to a weekend hackatjon to see if PEPFAR's numbers held up. What we found was that, yeah, there's a pretty strong case for PEPFAR. Even using conservative assumptions and ignoring many of its positive impacts, our best guess is that the program indeed saved 19million lives by 2018.

Going to get this in before the SB 1047 decision is made, so that it doesn't just come across as commentary on the decision: this was the California legislative process I've reported on most closely and it actually made me feel notably more proud of California democracy. The state politicians I talked to all seemed pretty smart, they'd talked to a lot of different people and could articulate several different perspectives, they did a good job of having their eye on the ball and seeing what the most important questions and disagreements were.

A bunch of people learned about Don Poldermans' use of fictitious data in the DECREASE trial series that informed European surgical guidelines from my newsletter last Friday. I got a bunch of questions, in particular about the estimated fatalities (🧵) And I wanted to talk a bit more baout whether Poldermans really killed 800,000 people. Here's what I wrote in the newsletter​:

Okay I'm sorry but I absolutely despise the 'elite hypocrisy' line here. No society has ever done more than ours to require poor people to live like the elite do, and this is often really bad for them. We ban cheap housing because it's better for people to live in nicer housing.

https://twitter.com/AltHistoryFan/status/1822492878238564508

We ban (as child neglect, for which the punishment is stochastic 'never seeing your child again') having your upper elementary school aged children walk home from school, let themselves in, and work on their homework until their parents get home. Hire a babysitter!

I think "there was a deal and it has broken down" is an incredibly powerful and pervasive sentiment in tech, not just among Trump supporters but among committed and sincere liberals too.

https://twitter.com/MattZeitlin/status/1813310972708458882

What was the deal? Hard to pin down exactly but something like - we will build ambitious things and pay high taxes and donate lots of money and mostly not play politics and you will treat us as valued pillars of our community, make our cities livable, stay mostly out of the way.

I was surprised by this, as the last official count I'd heard was around 35,000, so I clicked through to see what happened. What happened is that they argue that for every direct death in conflict there are often > 4 indirect deaths. So they multiplied the death toll by 4.

https://twitter.com/vijayprashad/status/1810104767869264173

I am worried this is not a very good methodology for estimating civilian deaths in Gaza. I had some trouble figuring out what they're citing for the rate of direct to indirect deaths in conflict zones, because the Lancet editorial links an unrelated UN pdf about the drug trade..

Scoop: OpenAI's senior leadership says they were unaware ex-employees who didn't sign departure docs were threatened with losing their vested equity. But their signatures on relevant documents (which Vox is now releasing) raise questions about whether they could have missed it. vox.com/future-perfect… Vox reviewed separation letters from multiple employees who left the company over the last five years. These letters state that employees have to sign within 60 days to retain their vested equity. The letters are signed by former VP Diane Yoon and general counsel Jason Kwon.

I'm getting two reactions to my piece about OpenAI's departure agreements: "that's normal!" (it is not; the other leading AI labs do not have similar policies) and "how is that legal?" It may not hold up in court, but here's how it works: OpenAI like most tech companies does salaries as a mix of equity and base salary. The equity is in the form of PPUs, 'Profit Participation Units'. You can look at a recent OpenAI offer and an explanation of PPUs here: levels.fyi/blog/openai-co…

When you leave OpenAI, you get an unpleasant surprise: a departure deal where if you don't sign a lifelong nondisparagement commitment, you lose all of your vested equity: vox.com/future-perfect… Equity is part of negotiated compensation; this is shares (worth a lot of $$) that the employees already earned over their tenure at OpenAI. And suddenly they're faced with a decision on a tight deadline: agree to a legally binding promise to never criticize OpenAI, or lose it.

You may have seen the story that GPT-4 told a taskrabbit it was blind in order to solve a captcha. The team that conducted safety testing, ARC evaluations, has a blog post out now about how that test went down: evals.alignment.org/blog/2023-03-1… The big things that confused me about the original story were: why was GPT-4 asking a Taskrabbit for help instead of using a service like 2Captcha? Which steps here did GPT-4 do independently? The blog post was helpful for explaining those things.

People might think Matt is overstating this but I literally heard it from NYT reporters at the time. There was a top-down decision that tech could not be covered positively, even when there was a true, newsworthy and positive story. I'd never heard anything like it.

https://twitter.com/mattyglesias/status/1588190763413868553

For the record, Vox has never told me that my coverage of something must be 'hard-hitting' or must be critical or must be positive, and if they did, I would quit. Internal culture can happen in more subtle ways but the thing the NYT did is not normal.

I have now had someone *impersonate a virologist* (badly) in order to, I assume, learn what emails I'm sending to schedule interviews about the Covid preprint claiming to find a synthetic lab origin. Setting aside that this is insane behavior, happy to show my work. When I learned that this preprint had been released, I read it, talked to my editor about it, decided it was worth digging into deeply to understand the analysis and any flaws in it and deliver a full explainer on the claims in the paper and whether they stand up to scrutiny.