The science of understanding AI.

Berkeley, CA
Why is interpretability of AI models important?
8
25
5,796
InterpretAI retweeted
Coming out: Encounters with AI and First Encounters with AI 😊
5
24
771
InterpretAI retweeted
And @kyliu99's essay The art of Copying from First Encounters with AI, recently published also in @LAReviewofBooks
Here's Ted on photography v writing :) in First Encounters with AI. His skepticism comes from the way LLMs work and are used - but he doesn't close the possibility of a different machine being more useful for art.
1
6
313
InterpretAI retweeted
If AI will do your thing, you need to find something crazier that AI wouldn't think of. Easy.
5
17
1,015
InterpretAI retweeted
What is college if knowledge, teaching, and advising are available from machines? One model has answers to many challenges that AI is posing today: residential education where students learn from each other to be bold, original, and uniquely human while living, eating and socializing together. begus.substack.com/p/the-fut…
1
4
19
1,041
InterpretAI retweeted
Handout from today. zenodo.org/records/22803280
7
18
1,401
InterpretAI retweeted
My talk on a new conceptualization of language at the Science of Intelligence seminar at Berkeley. archive.org/details/redwood-…
10
40
3,465
InterpretAI retweeted
My coping mechanism in the age of AI: write more Proto-Indo-European papers. Here’s a story on the origin of PIE mobility. I argue that in Pre-PIE, a rule operated that whereby heavy syllables attracted stress from the root to the ending. I focus on root nouns and verbs.
4
21
93
6,224
i often think doing a dissertation on james joyce at oxford was an early stage in my NLP/LLM work….
Yes, Joyce was exploring the prior layers of language. Here's a project that lets you explore those prior stages: latentspacecraft.antikythera… Glad to se that the industry is coming to the same conclusion!
1
5
12
882
InterpretAI retweeted
Yes, Joyce was exploring the prior layers of language. Here's a project that lets you explore those prior stages: latentspacecraft.antikythera… Glad to se that the industry is coming to the same conclusion!
7
22
4,677
InterpretAI retweeted
A new book series just dropped!
7
12
957
InterpretAI retweeted
Best way to escape AI: start writing in cuneiform. It’s how it all started!
5
7
22
1,838
InterpretAI retweeted
Philosophy is coming alive again through frontier science and tech. I think we should do literally everything possible to support this energy and launch an intellectual and cultural renaissance in America. bury inherited boring tropes in waves of strange, curious, ambitious inquiry and experiment
1
7
46
1,308
InterpretAI retweeted
Two to three more Millenium problem are rumored to be solved. It's time to switch to questions with no provable answers! Academia will flourish when we go back to our roots. Meet George of Trebizond. A 15th century scholar who wrote endless pages thrashing Plato and revering Aristotle. He had a great life in Rome.
3
7
638
InterpretAI retweeted
AI will unite the biological world.
1
5
14
706
InterpretAI retweeted
Siblings are finally united! What is writing and reading in the time of AI? Ted Chiang, Ken Liu, Sheila Heti, Sasha Stiles, James Yu, and other amazing writers reflect and offer new possibilities.
3
7
923
InterpretAI retweeted
Two books walk into the latent spaces exhibition…
When a paper becomes a piece of art. We’re exhibiting our Latent Spacecraft GAN models with @mthvn @begusgasper at the Hidden Layers exhibition. Come see the artworks at the Worth Ryder Gallery in Berkeley. The opening reception is today, 9/3, and we run until 10/3.
2
4
14
882
InterpretAI retweeted
Siblings are finally united! What is writing and reading in the time of AI? Ted Chiang, Ken Liu, Sheila Heti, Sasha Stiles, James Yu, and other amazing writers reflect and offer new possibilities.
2
9
44
3,438
InterpretAI retweeted
Everyone can access the paper and the models online: latentspacecraft.antikythera… Model 1: a speech-GAN model that learned on 8 words of English latentspacecraft.antikythera… Model 2: a speech model that learned on Finnegans Wake only latentspacecraft.antikythera…
When a paper becomes a piece of art. We’re exhibiting our Latent Spacecraft GAN models with @mthvn @begusgasper at the Hidden Layers exhibition. Come see the artworks at the Worth Ryder Gallery in Berkeley. The opening reception is today, 9/3, and we run until 10/3.
7
23
1,356
InterpretAI retweeted
What are narrative agents? A new piece on the artificial humanities observatory Deckard by Fabrizio Allione and Riccardo Milanesi.
1
5
13
1,025
InterpretAI retweeted
In the Daily Cal today on the use of AI in the university’s public communications
2
5
21
2,265