Twitter Beğeni Hilesi

Incredibly proud of what this team has accomplished in just over a year since our launch. Built for the enterprise, our platform is backed by world-class research and operates in production at scale.

Don't forget to speak to Yintao Tai today about PIXAR at #ACL2024NLP during the "Findings In-Person Poster Session 2". Time: 17:45 (Bangkok time) Location: Convention Center A1 Feel free to send me a message if you would like to chat about Embodied AI and multimodal AI!

This is what I mean when I talk about input perturbations that completely break current LLMs: x.com/danielhanchen/… We are clearly missing something right? Pixel-based LLMs might be the answer!

Congrats to our team for winning two paper awards at #ACL2024! OLMo won the Best Theme Paper award, and Dolma won a Best Resource Paper award! All the credit goes to the whole team for the massive group effort 🎉🎉

It's so refreshing to see this paper being awarded. I hope that somebody will take on the challenge of making multimodal models as open as OLMo ;)

New on Air Street Press: Now that LLMs can convincingly automate much of a bored human’s tasks, attention is turning to “agentic AI”. In this piece, we evaluate how far advanced this work actually is, look at both promising research directions and the challenges ahead. Thread:

A good reminder for those of us in LLM land (like me) that we don't only need to mitigate gender biases *caused* by LM generation, but we should enable researchers to *use* LMs to discover biases in human content. From Isabelle Augenstein's keynote genderbiasnlp #ACL2024

If you're attending the #ALVR workshop today at #ACL2024NLP, go and read Malvina's work on Continual Learning for VLMs!

Ok, whatever it is that OpenAI has done to o1, it has payed off. At least on wordle, which used to be one of the hardest parts of our “conversational agency” benchmark. 4o: 23 (previous best) o1: 75.33 (Human expert players: 72)