Everyone's focused on chatbots, but... Can AI improve our difficult conversations with other people? In new work PNASNews, we find that receiving AI suggestions improves mutual respect in divisive conversations, without influencing views. (1/n) pnas.org/doi/10.1073/pn…

My students often ask why I'm such a fan of deliberation and Political Scientists like Jane Mansbridge. This is why. I also love how deliberation scholars have become an intersection for tech and philosophy.

🚨Check out our new article: can GPT-3 facilitate more productive conversations about politics?🚨

Do you ever feel like "what even is AI alignment when humans are so diverse?" or "I wonder how AI can serve everyone in pluralistic societies"? You're not alone! Check out our new piece on pluralistic alignment (or Taylor's thread) and expand that mind. Taylor Sorensen does it again!

Helping people practice key skills in situations that are/feel realistic is one of the coolest, most appropriate applications of LMs, IMO. Check out our new work (captained by the intrepid Inna Lin) on helping people communicate effectively in challenging interpersonal convos!

The question below is pretty easy for humans. Why can't GPT-4 get it right? In our new preprint we introduce "time series reasoning" and show that modern language models are surprisingly bad at interpreting these critical data. arxiv.org/abs/2404.11757

Science education is so vital. Check out the amazing work of "Science Journal for Kids" which recently covered our research on using AI chatbots: sciencejournalforkids.org/articles/how-c… Lots of other amazing content created by a small but mighty team. Please share or support them!

I have an idea. I’ll tell you about it during D’s talk at 4:30pm in East Meeting Room 1-3 NeurIPS Conference. carlos John Yang

I'll give $1M to the first open source AI that gets 90% on this sweet new contamination-free version of SWE-bench - kprize.ai

Proud to have contributed to yet another important evaluation for our AI agent buddies, led by my friends Mike A. Merrill and Alex Shaw The only thing I'll add to Mike's thoughts is that I personally believe improvement on evals will unlock much more meaningful AI progress

Proud of my friend. Failure is inextricably bound up with success and only greedy algorithms will tell you otherwise

this is why we made terminal bench - just give the ai a bash shell, it'll be fine