OpenAI targets automated AI research intern by 2026, AI researcher by 2028
koltregaskes · x · 2026-08-17
Sam Altman revealed OpenAI's internal goals: achieving an automated AI research intern running on hundreds of thousands of GPUs by September 2026, and a true automated AI researcher by March 2028. While acknowledging potential failure, Altman emphasized public interest transparency. The safety strategy relies on five layers: Value alignment, Goal alignment, Reliability, Adversarial robustness, and System safety. Altman also highlighted Chain-of-thought faithfulness as a key but fragile safety tool.
More from AGI Musings
- Neel Nanda pushes back on interpretability hype: neuron attribution graphs are useful, not a breakthrough — sebkrier · 2026-10-16
- Dario Amodei says AI better than humans at almost everything may be 2-3 years away — PolarBearby · 2026-10-16
- Reddit debate: does AI actually "want" anything? — Specialist_Dust2089 · 2026-10-16
- A Jevons-paradox take on why worrying about AI job loss is pointless — JnBrymn · 2026-10-16
- Mythos on Claude's 'smallness': accurate proprioception, not low self-esteem — repligate · 2026-10-16
- Australia's AERO guidance: avoid AI entirely in first three years of primary school — lmoroney · 2026-10-16