also, people are much better at prompting nowadays

Hot take: LLMs haven’t gotten much smarter at coding in the last two years. What makes the difference are larger context windows, context engineering, and function calling.

Everyone suddenly got promoted from prompt engineer to context engineer

From now on I will end every query to a transformer model with “Thank you for your attention to this matter!”

It’s so funny having @garymarcus.bsky.social speak at the Web Summit opening ceremony. All the hustlers here are pumped about AGI and agents and Gary is like nah it’s a bubble. Great start!

The best thing about mornings

Maybe the goal was never to beat ARC-AGI, but rather the friendships we built along the way

From Ilya’s NeurIPS talk today: pretraining as we know it will end. We have but one internet. The internet is the fossil fiel of AI.

LSTMs are apparently considered an “ancient history” today. 🤣

NeurIPS 2024 Pack Join the conversation

I just updated the NeurIPS starter pack with many more attendees

Let me know if you'd like to be added

go.bsky.app/BuJXg5q

#NeurIPS2024 #NeurIPS

Someone made a starter pack with folks at NeurIPS this year

bsky.app/starter-pack...

You can finally know if you had a reason to complain when your paper got rejected

Have you shipped any LLM-powered apps in production? Which LLMs do you use? I am curious if people actually deploy open-source models for production apps or default to readily available APIs from OpenAI / Anthropic / Google

You know NEURIPS is off to a good start when the opening remarks have a slide with a Ted Chiang story

Added some more folks to the Open Source AI Starter Pack:

go.bsky.app/N8yVZdW

Posting a call for help: does anyone know of a good way to simultaneously treat both POTS and Ménière’s disease? Please contact me if you’re either a clinician with experience doing this or a patient who has found a good solution. Context in thread

Screenshot of how to call Gemini through openai sdk.

I just learned that Gemini is supported in the openai sdk. It also supports structured outputs 🤷

*image originally shared by logan Kilpatrick on the ex-bird app

Are there any independent researchers (not affiliated with universities) here? I don't want to quit my day job and move to academia, but I am considering exploring a few AI research ideas in my free time and publishing a paper. Has anyone tried that?

Hi everyone! 👋 How do I find AI / ML / Data Science folks here?