Sam Altman reveals exact date of intelligence explosion

OpenAI projects achieving an AI research intern by September 2026 and a legitimate AI researcher by March 2028, leading to an intelligence explosion shortly after.

• OpenAI predicts an 'intern-level AI' by 2026 and a 'legitimate AI researcher' by March 2028, leading to an 'intelligence explosion.'

• Company is heavily investing in a multi-trillion-dollar infrastructure build-out and exploring 'chain of thought faithfulness' for AI safety.

• OpenAI's new structure includes a nonprofit governing a public benefit corporation, with significant commitments to AI for health and resilience.

OpenAI's Sam Altman and Jakub Pachaki shared a precise timeline for AI development, targeting an AI research intern by September 2026 and a full AI researcher by March 2028. This accelerated development is expected to lead to an 'intelligence explosion' and subsequent superintelligence, with the company heavily investing in infrastructure and addressing safety concerns like AI addictiveness and faithfulness in its models.

OpenAI's AI Timeline

• 00:00:00 OpenAI anticipates developing an 'intern-level AI research assistant' by September 2026 and a 'legitimate AI researcher' by March 2028. This timeline suggests a rapid acceleration of AI capabilities, leading to an 'intelligence explosion' where AI improvement becomes recursive and limited only by available compute, quickly reaching superintelligence. The company's core research program is focused on achieving this self-improving artificial intelligence, which is seen as a critical race among frontier labs.

AI Safety and Faithfulness

• 00:05:13 OpenAI is researching 'chain of thought faithfulness' to ensure models are aligned with human incentives and transparently reflect their internal reasoning without human supervision during training. This approach, which involves allowing models to 'think' independently, is believed to provide greater insight into their processes and enhance safety, though its application requires careful boundaries and abstraction to avoid fragility. Sam Altman also expressed concern about AI's potential for addictiveness, pledging to withdraw problematic products like Sora if they become overly addictive.

Ambitious Infrastructure Plans

• 00:09:36 OpenAI is pursuing an enormous infrastructure expansion, with 1.4 trillion dollars already committed towards an initial 7 trillion dollar 'Stargate' plan for AI infrastructure. The company aims to build 'AI factories' capable of producing gigawatts of compute per week, indicating a future where the limitation on AI acceleration will be compute power rather than duration of autonomous tasks. Robotics are also envisioned to play a key role in building these data centers.

OpenAI's New Corporate Structure

• 00:11:00 OpenAI has finalized a simpler corporate structure consisting of the OpenAI Foundation (a nonprofit) and the OpenAI Group (a public benefit corporation, or PBC). The nonprofit governs the PBC, holding 26% equity and potentially more in the future, with the PBC attracting resources for fundraising, including an eventual IPO. The OpenAI Foundation has also pledged $25 billion towards AI health, disease curing, and AI resilience initiatives.