I still can't get over the fact that the psychopath behind the "protests" blocking aid to Gaza is an American who fled to Israel to escape charges for his suspected role in the murder of Alex Odeh, a Palestinian in the US. @JoeBiden could literally extradite him and stop this. Doing so would simultaneously help with justice for Alex Odeh and aid for Gaza. How hard can it be to extradite him? He is an American who is already wanted for murder in an act of domestic terrorism. Extradite Andy Green, @JoeBiden!

I think few people know what this symbolism is, so it's worth noting that this is the flag of Kach, a Kahanist symbol. Kahanism is the genocidal ideology of Itamar Ben-Gvir and his ilk. It strives for a solely Jewish state from the river to the sea, and views all Arabs as enemies Wikipedia article on Kahanism: en.wikipedia.org/wiki/Kahanism

So I was in Morocco for the past two weeks. I didn't say anything about it publicly because of the Level 3 travel warning for Israeli citizens, and my very public Israeli-American identity and many posts about Israel/Palestine the last few months. It was amazing, I love Morocco I'll post photos and details soon. The coolest thing I did was eat in Maimonides' house in Fez. With my boyfriend's Muslim family! But a lot of my posts the past two weeks will make more sense knowing I was in Morocco lol

Not just saw---it was. The Soviets who liberated death camps just gave survivors a few pennies and a horse and sent them on their ways. My grandpa said many died shortly after liberation by the Soviets from the abrupt transition

https://twitter.com/AntifaCatraa/status/1737348219083141384

And then they came back to see their property and wealth stolen, and their entire extended families wiped out. Those who tried to repatriate often faced extreme violence in response, especially in Eastern Europe

This is one of these things that is hard to say tactfully, but I think it's really important, so I will try anyways; please be kind. On Israeli-Palestinian friendship, or even plain old conversation🧵: I never met a Palestinian for the first several decades of my life. Nor spoke to a Palestinian online.

The reason I care so much about folks from MENA countries outside of Palestine acknowledging the ethnic cleansing of Jews from those countries is not because it changes anything about Israel/Palestine, but rather because it changes the power dynamic of these conversations.

https://twitter.com/TaliaRinger/status/1732130080401375720

It's one reason why I have less trouble talking to Palestinians about this than talking to folks from other MENA countries who deny or minimize this ethnic cleansing, or blame it on Israel.

We just learned that our paper Baldur: Whole Proof Generation and Repair with Large Language Models got a Distinguished Paper Award at ESEC/FSE! Emily First is first author, with Markus Rabe, me, and Yuriy Brun. Come to Emily's talk next Wednesday!

2023.esec-fse.org/details/fse-20… There is a preprint here:

Will post the published version once it is availablearxiv.org/abs/2303.04910

Terry Tao and I spoke over coffee for like two hours yesterday, in part about diversity in how people think about math. We both agreed that people who hit these walls early mostly don't learn the way of thinking about math that works for them. It's an educational failure

https://twitter.com/charlesmurray/status/1680856124315123712

The educational system for some reason gives people the impression that if they cannot think about math in the way they are taught to think about it, they are just bad at math. But often they just need a different way of thinking about it

Am I supposed to be able to hear anything in this xAI Twitter Space thing I'm just here to hear updates from @ChrSzegedy and @Yuhu_ai_ 😭

I really do think the idea of machines becoming superhuman on all "cognitive" tasks is just a fiction. Every time we develop machines that outperform us on some task, we find ways to build on that to find new tasks. We have superhuman calculators already I think the cool thing about people is our infinite adaptability, and I don't think that will stop no matter what things we care about now that machines get good at. We'll just care about new things

When a student is really stuck on a proof of a top-level theorem, one way we figure out what lemmas we need works as follows: (1) The student picks particular example inputs. So if the theorem says "forall x," the student chooses a few examples of x that cover different cases.

Imagine thinking that those of us who disagree are necessarily not experts. There are so many experts who don't agree with anything about this, and many who don't even think "AGI" is a meaningful concept. I mean, hell, even within the same Turing Award you've got one holdout

https://twitter.com/Jabaluck/status/1663744727664013316

Anyways, hello, I am on a National Academies committee for AI for Mathematics, and I think "AGI" isn't a consistent and well-defined concept worth discussing to begin with, especially within the realm of science. Mostly, I think it is racialized pseudoscience dressed up for 2023.

There are some days when my emotional empathy goes so far into overdrive that it feels like I can feel everyone's feelings all at once, and I can't tell which feelings are mine versus theirs. Also, everyone looks extremely familiar It feels like a form of fatigue, like I am constantly working very hard to delineate between myself and others, and between the familiar and the unfamiliar, so that I do not take on everyone's feelings at once. But sometimes I get too tired to maintain that boundary

ah yes, the six genders Which is less sweet

The weirdest thing to me about grad school was learning that there is a whole portion of the US population that goes to ultra-elite high schools, and from there to elite universities, and from there to elite PhD programs, and from there to faculty life.

https://twitter.com/constans/status/1663597659335032843

And for these people, that's the default path. It just never occurred to me that those people existed before. It makes sense now. But I had no contact with them growing up, except at Japan Bowl Nationals when they got first place.

Proposal to center genocide in the conversation about AI Risk, both in terms of (1) stopping and repairing harms of ongoing AI-aided genocides, and (2) preventing future AI-aided genocides.

https://twitter.com/TaliaRinger/status/1663591127776219136

Reason one: Many of us do not think that extinction is a worse outcome than genocide. To make that inference requires some moral assumptions many of us who instinctively react negatively to the entire x-risk movement do not hold.

https://twitter.com/TaliaRinger/status/1663591438574075906?t=BEZMTu5IfSS84KXng70tpQ&s=19

I think I am getting a lot of new followers who are interested in mechanistic interpretability because of our latest draft. Cool, hi! But also, you should know I am unlikely to have palatable opinions on "AGI" to you (I believe the concept is harmful to entertain) So, we are most likely to get along well on research topics if we avoid discussing "AGI" and so on, and just focus on concrete immediate-term problems and solutions, which is where I think mechanistic interpretability is extremely useful

New preprint just dropped! "Can Transformers Learn to Solve Problems Recursively?" With @dylanszzhang, @CurtTigges, @BlancheMinerva, @mraginsky, and @TaliaRinger. arxiv.org/abs/2305.14699 Recursion is important for a lot of programming languages and formal reasoning tasks. We wanted to know to what degree transformer models could learn to approximate interesting recursive functions from I/O examples---and how they would approximate them, and how they would fail!

When open source and industry folks are establishing working relationships beyond a short summer stint, they should look toward student-professor pairs rather than just students or just professors. (Postdocs alone make sense though) If you establish a relationship with just a professor, but not any of the professor's students, then there is a high chance that the professor abandons that line of work once they have their own students to work with. It's like working with an industry manager but not their team

An undergrad was asking me about grad schools to apply to, and had been advised to apply to Oxford by another letter writer. My response was that Oxford is "basically eugenics central" these days. Would not advise

https://twitter.com/timnitGebru/status/1661186282721083392

And I'm sure there is great work at Oxford, but that doesn't make me feel like I could in good faith recommend a student go to a place where that kind of nonsense is so mainstream and Bostrom is in a position of any power. I feel somewhat similar about Stanford, honestly

Just want to announce that I've been building a god-machine for profit by way of using your data without acknowledgement or pay. This is a good thing, because I am a Messiah chosen for the holy task of doing this responsibly In fact, it is urgent that I steal your data to build a god-machine for profit immediately, because if I were to not do that, someone else would. And they would be worse