🚨Exciting new findings from our lab!

This work challenges the concept of 'Zero-Shot Generalization' in multimodal models.

Contrary to being 'zero-shot', these models actually require exponentially more data to achieve only linear improvements in perf.🔍🤔!

Detailed 🧵👇

Widespread availability of Gemini 1.5 Pro in 180+ countries, with long context and audio capabilities. Now in public preview.

🔥🔥

I'm not surprised at all -- I predicted this back in 2016! This is so well deserved, congratulations!

Despite impressive advances in AI and robotics, we are nowhere near humanoid robots like Rosie. andréa graziano and I explored this fundamental challenge with Dileep George from Google DeepMind, Giacomo Indiveri | @[email protected] from ETH Zurich and Stanislaw Wozniak from IBM.

youtube.com/watch?v=KIKweX…

Our research project SIMA is creating a general, natural language instructable, multi 3D game-playing AI agent. The agent can carry out a wide range of tasks in virtual worlds, making AI more adaptable, helpful & fun! dpmd.ai/sima-1

Gary Marcus Petulant Barbarian Kashif Ansari The New Yorker Have fun :-) blog.dileeplearning.com/p/welcome-to-t…

Timothy B. Lee Dileep George has a great post summarizing this view blog.dileeplearning.com/p/ingredients-…

Went to LensCrafters to get my prescription updated…the doctor told me she got 20 other folks doing that today just like me for getting the Apple Vision Pro 😳

Congrats to the #AlphaGeometry team on their impressive breakthrough published today in nature using a neuro-symbolic hybrid AI system that can solve Maths Olympiad-level geometry problems. deepmind.google/discover/blog/…

If animals could speak they’d tell us that LLMs are dumb 😇

Only sweet lessons in this book 😇

I've studied intelligence all my long life, yet still I feel I learned important things about intelligence by reading this book.
Thank you, Max Bennett.

Max Max Bennett says the brain's major evolutionary changes are (to a first approximation) all about discovering new sources of learning. We take a pleasant stroll through the 5 breakthroughs he details in his book A Brief History of Intelligence.

braininspired.co/podcast/181/

enjoyed this from Dileep George -- are transformer models like dirigibles?

blog.dileeplearning.com/p/welcome-to-t…

Oops

Excited for our latest breakthrough published in nature today, #FunSearch ('searching' in 'function' space) is the first system to make new verifiable discoveries in open problems in math & CS, we applied it to find a more efficient solution to the classic 'bin packing' problem

6/ At Fifty Years we're optimistic about tech uplifting prosperity for all. The Progress Index sheds light on untapped opportunities. Check it out: progress.fiftyyears.com

1/ Crazy question - ever think about making paper?

Too many startups are focused on progress in the same few areas like SaaS. Brilliant people who could build massive leaps forward instead grind out incremental features in the same tired spaces.

Most webscrape data sets were O(10T) tokens. That represents most of all human content ever created.

If GPT4 made $100m at $20/million tokens, that’s 5T tokens.

See the problem we’re facing ahead of us?