Algorithms vs Compute: A Game-Theoretic View on AI Safety
kellerjordan0 · x · 2026-08-11
The post explores the inherent contradictions in AI safety from a sci-fi perspective:
- Algorithms researchers argue for pushing algorithmic progress as fast as possible to see if current compute levels are already dangerous, thereby necessitating a freeze on compute production.
- Hardware engineers argue the reverse: build compute as fast as possible to reveal if current algorithms are too dangerous, necessitating a freeze on algorithmic research.
The author concludes that the winner is whoever's form of progress would be harder to freeze, guessing it's probably the algorithms researcher.
Related event: The AI Safety Paradox: Algorithmic Progress vs. Compute Expansion(4 posts)→
More from AGI Musings
- Tech Giants' Compute Budgets Eclipse US Federal Spending—We Live in Cyberpunk Now — tszzl · 2026-08-11
- Researcher: LLMs Ruthlessly Deconstruct Flawed Empirical Social Science Papers — RexDouglass · 2026-08-11
- Jim Chanos Hints AI Bubble Will Burst in Congressional Hearings — zephyr_z9 · 2026-08-11
- Do LLMs Truly Generalize? Skepticism Arises Over the Scaling Law Myth — JacquesThibs · 2026-08-11
- Gary Marcus Reiterates Warning: LLMs Are Still 'Bulls in a China Shop' — GaryMarcus · 2026-08-11
- Looking Back at 2015: AI Experts' Top Risk Predictions for the Next 20 Years — danfaggella · 2026-08-11