2024-09-24T04:43:33+00:00 | 🔗
It’s hard to skip the ad when your mid lawn-mow or mid dish-wash. https://t.co/kjGgkaEr2f
2024-09-23T21:52:16+00:00 | 🔗
Models/architectures like o1 are usefully because they can ground **highly-complex evals** which can then be used the prompt engineer and optimized LLM pipelines.
2024-09-23T20:03:01+00:00 | 🔗
Perfect example of how "intelligence" and "economic value" are so poorly defined and not highly correlated. https://t.co/ByurNYp4U5
2024-09-23T19:45:34+00:00 | 🔗
And he is right! https://t.co/nbemtkxeNB
2024-09-23T19:42:11+00:00 | 🔗
@aashaysanghvi_ I think that GPT-5/ GPT-6 will have the same increase in real world impact as GPT-4 vs. GPT-3 and GPT-3 vs. GPT-2, namely incremental. I don't think that there are any applications that require leaps in foundational models of language -- only ways to multi-agent architect models.
2024-09-23T18:08:28+00:00 | 🔗
Persi Diaconis got me to spend weeks practicing the Faro (which I still can't do with high probability) https://t.co/njitR6C20x
2024-09-23T04:10:16+00:00 | 🔗
@emollick Paper from Google Research on auctions within RAG https://t.co/Mt1IU76N7L
2024-09-23T04:03:08+00:00 | 🔗
@GrantSlatton https://t.co/SjbHWSk0Os "important to the decipherment of cuneiform, as it is the longest known trilingual cuneiform inscription, written in Old Persian, Elamite, and Babylonian (a variety of Akkadian)"
2024-09-23T02:55:45+00:00 | 🔗
“All religions are a path to arrive at God,” he said. “They are like different languages to arrive there. But God is God for all.” https://t.co/TVA6t7Smyw
2024-09-23T02:46:12+00:00 | 🔗
Things can change on a dime when the impediments weren't real issues! https://t.co/T5WbD7tIkU