BREAKING
Sutton Launches Oak Lab
A Bet Against Scaling
Two Paths to AI
Scaling-first LLMs
Today
●
Ever-larger models
●
Static crawled datasets
●
Excel at imitation
Oak Lab
New
●
Continual learning
●
Runtime experience
●
Builds world models
0
T
parameters
0
W
energy budget
0
Turing Award
How Oak Learns
1
Learn in real time
↓
2
Build world models
↓
3
Self-evaluate
↓
4
Select and adapt
No Prototypes Yet
AI NEWS BLITZ
Turing Award winner Richard Sutton has launched a new AI lab called Oak Lab.