Google reportedly pre-builds compute for AI self-improvement, with RSI eyed for 2027–2028
kimmonismus · x · 2026-08-04
Google is reportedly betting that AI will help build better AI, with DeepMind strategy chief Jasjeet Sekhon calling recursive self-improvement a key part of the investment thesis.
- The company is pre-building compute for a possible discontinuity in AI R&D.
- Sekhon said current AI revenues are not yet enough to fund the capex.
- DeepMind and OpenAI researchers reportedly place the timeline for such a jump around 2027–2028.
- The attached clipping also quotes DeepMind’s Oriol Vinyals and OpenAI co-founder Wojciech Zaremba saying RSI might be achieved by 2027 or 2028.
More from AGI Musings
- 90% of People Haven't Used AI Agents, Creating a Massive Capability Divide — Aizkmusic · 2026-08-04
- 2024: AI that talks, 2026: AI worth talking to — deepgramscott · 2026-08-04
- NVIDIA, Meta, and 25 Others Sign Open Letter Backing Open Weights for US AI Leadership — hugobowne · 2026-08-04
- A proposed AI ethics rule centers on maximizing total human autonomy under truth — GlenBradley · 2026-08-04
- AI ethics should be one first-principles rule embedded across the whole model stack — GlenBradley · 2026-08-04
- Peter Diamandis says AI benchmarks may matter less than real-world prediction — PeterDiamandis · 2026-08-04