Jeff Dean
@JeffDean · Nov 7, 2025
An exciting new approach for doing continual learning, using nested optimization for enhancing long context processing.
Google Research@GoogleResearch· Nov 7, 2025Introducing Nested Learning: A new ML paradigm for continual learning that views models as nested optimization problems to enhance long context processing. Our proof-of-concept model, Hope, shows improved performance in language modeling. Learn more:
Elon Musk
@elonmusk
Not sure if it’s super new, as we’ve been doing something along these lines at Tesla for a while, but it does make sense.
The single biggest technical challenge of Tesla self-driving AI is context compression of 1.5GB/s of video to ~2kB/s of control outputs using a puny inference computer without making any mistakes.
Super hard to avoid overly lossy compression at any given step.
08:31 AM · November 8, 2025 · 1.1M views
164
209
2.8K