Justin Joque

Bring back Derrida!

nytimes.com

oh man I feel real bad for Anthropic, it really sucks when you work super hard on something and then someone just takes it without permission to train an LLM

Chinese AI labs secretly used millions of Claude exchanges to train their models, Anthropic says cnbc.com

Some of this argument is here in this short review and also in my next book which is nearly done olrsupplement.com/2026/02/17/t...

The AI Con, by Emily Bender and Alex Hanna - The OLR Supplement olrsupplement.com

In my AI class today I'm going to try to convince my students that super intelligence has already arrived, far surpassed humanity, already rules the world with humans doing its will and it is not a computer, it is corn

Holy shit you all. A friend who works in admissions at a school that has a decent number of online program was telling me they're having problems with identity thieves using a stolen identity to apply, get in, and get someone else's financial aid. That's been happening for a while but now with AI...

I'm bored of AI solving all these silly number theory problems, let's see it decrypt the Voynich manuscript

him: AI is not conscious and will never replicate human intelligence me: unlike AI the human being essentially occurs in his essence only where he is claimed by being. Only from this dwelling does he "have" "language" as the home that preserves the ecstatic for his essence. We are not the same

The interesting question about large language models is not whether or not they are sentient it is whether there could be any way to buy them for their exchange value yet exploit them for their use value (hence allowing them to create surplus value, a feat so far limited to humans)

until AI can sit in a 190F room, throw water on the rocks and talk about French New Wave films, I think the humanities will be just fine.

Me: I'm so sick of the repetitive AI discourse. I wonder what the esteemed subjects of twitter discourse of the days of yore, Merve Emre and Agnes Collard are up to....well would you look at that

I'm not like an AI architect or anything but maybe they should put a secret calculator in chatgpt. Seems a lot more efficient than reasoning 457 * 58 from first principles. As a human I'd also be like "I know this is a problem for calculator"

Any analysis of artificial intelligence should start neither with a particular software nor with the promise of AGI but rather with what I would call General Artificial Intelligence. This would name the entire constellation of softwares, hardware, venture capitalists, weird Harry Potter sex cults...

I asked AI to peer review my children's contributions to our halloween decorations and bad news for them none of them made the cut. Revise and resubmit.