BREAKING
Stanford Lecture Resurfaces
0
OpenAI speakers
Why Language Models Scale
1
Predict tokens
↓
2
Learn many tasks
↓
3
Gain capabilities
Transformer History as a Guide
1
Cheaper compute
↓
2
Larger training
↓
3
Design shifts
Useful, but Often Mislabelled
Documented content
CS25
●
Language-model intuition
●
Scaling and Transformer history
Unsupported framing
Caution
●
New agent tutorial
●
Workforce and revenue claims
Foundation Models, Not Agents
AI NEWS BLITZ
A Stanford CS25 session has found a new audience more than two years later.