# Jack Rae

Markdown copy of https://realalpha.app/builders/jack-rae/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

Real Alpha has cut 5 moments of Jack Rae (Distinguished scientist, Meta) talking, from an interview on Unsupervised Learning published Mar 17, 2025.

- Role: Distinguished scientist, Meta
- About: Led Gemini Thinking, co-led Gemini pre-training, and contributed to Gopher and Chinchilla language-model scaling research.
- Company: [Meta](https://realalpha.app/companies/#meta)
- On X: https://x.com/jack_w_rae

## What Jack Rae said

- [AI math progress depends on asking better questions](https://realalpha.app/s/ai-math-progress-depends-on-asking-better-questions/)
  Unsupervised Learning · Mar 17, 2025 · 6 min
  > “Whether it's possible, I'm not sure, but I would agree maybe the key crux is can I ask actually novel questions?” — Jack Rae · [31:30](https://realalpha.app/v/atMRWzgHEGg/?t=1890)
- [Qualitative tasks need broad reward criteria](https://realalpha.app/s/qualitative-tasks-need-broad-reward-criteria/)
  Unsupervised Learning · Mar 17, 2025 · 2 min
  > “take in a very broad set of criteria and then apply a reward signal once we have the reward signal we can train with reinforcement learning against it” — Jack Rae · [10:54](https://realalpha.app/v/atMRWzgHEGg/?t=654)
- [Agent research depends on choosing the right environment](https://realalpha.app/s/agent-research-depends-on-choosing-the-right-environment/)
  Unsupervised Learning · Mar 17, 2025 · 4 min
  > “I don't think anyone has solved the perfect environment yet” — Jack Rae · [18:57](https://realalpha.app/v/atMRWzgHEGg/?t=1137)
- [Use one model unless specialization earns its cost](https://realalpha.app/s/use-one-model-unless-specialization-earns-its-cost/)
  Unsupervised Learning · Mar 17, 2025 · 2 min
  > “If there's positive transfer, put it in the same model, as long as it doesn't then become too expensive to serve.” — Jack Rae · [43:55](https://realalpha.app/v/atMRWzgHEGg/?t=2635)
- [Inference-heavy AI can use more distributed compute](https://realalpha.app/s/inference-heavy-ai-can-use-more-distributed-compute/)
  Unsupervised Learning · Mar 17, 2025 · 3 min
  > “if it turns out that building AGI becomes mostly an inference problem, an inference problem that can be much more distributed than maybe synchronous large batch training” — Jack Rae · [1:06:44](https://realalpha.app/v/atMRWzgHEGg/?t=4004)

## The episode

- [Noam Shazeer and Jack Rae: Scaling Test-time Compute, Reactions to Ilya & AGI](https://realalpha.app/v/atMRWzgHEGg/) · Unsupervised Learning · Mar 17, 2025