MemCast
MemCast / episode / insight
Seven factors determine AI progress more than algorithmic cleverness
  • Raw compute
  • Data quantity
  • Data quality/distribution
  • Training duration
  • Scalable objective functions (pre-training, RL)
  • Numerical stability
  • Normalization techniques
  • These factors explain most progress since 2017
Dario AmodeiDwarkesh Patel00:03:13

Supporting quotes

One is how much raw compute you have. The second is the quantity of data. The third is the quality and distribution of data. It needs to be a broad distribution. The fourth is how long you train for. Dario Amodei
Then the sixth and seventh were things around normalization or conditioning, just getting the numerical stability so that the big blob of compute flows in this laminar way instead of running into problems. Dario Amodei

From this concept

The Scaling Hypothesis

Amodei explains his 'Big Blob of Compute' theory - that AI progress depends primarily on seven scalable factors rather than algorithmic breakthroughs. He discusses how this applies to RL and generalization.

View full episode →

Similar insights