Interview
Sam Altman: OpenAI, GPT-5, Sora, Board Saga, Elon Musk, Ilya, Power & AGI | Lex Fridman Podcast #419
- Compute is projected to become the world's most precious commodity, with future demand described as a "tremendous amount" driven by a pricing model where usage scales to the cost X.
- By the end of the current decade, possibly sooner, "quite capable systems" viewed as "really remarkable" are expected to emerge, with the gap between GPT-5 and GPT-4 anticipated to mirror the previous leap between GPT-4 and GPT-3.
- A "new paradigm" for slower, sequential thinking is foreseen, alongside a strategy of iterative deployment to allow societal adaptation, while OpenAI plans to release an "amazing new model" this year with an unspecified name.
- Context length capabilities are expected to eventually reach "several billion" in the near future and feel "infinite" or "trillions upon trillions" in the distant future, with Sora and similar models predicted to improve continuously.
- Nuclear energy strategies include a re-embrace of nuclear fission for new reactors, reliance on Helion for nuclear fusion, and a continued future return to robotics.
- In the next five years, AI tools are expected to produce "many, maybe most," YouTube videos, which will remain driven by human thought despite AI involvement in production.
- The path to Artificial General Intelligence (AGI) is expected to involve a "giant power struggle," an "arms race" increase in competition, and increasing political polarization between "left versus right wars."
- Risks include "bad ones" among AI consequences, a non-zero chance of surprises regarding existential risk, and the potential for increased anxiety about a faster takeoff, whereas a "slow timeline to the start of AGI" is identified as the safest and most preferred outcome.
- OpenAI anticipates facing "more and more" pressure as it approaches AGI, with governance evolving to avoid any single person or company holding super voting control to make these decisions alone.
- Internal expectations include continued collaboration with Ilya Sutskever for the remainder of Altman's career, with Sutskever focused on the 10-year horizon, while the new board is expected to foster a more organized and less dramatic operating environment.
- Future interpersonal dynamics are expected to see an "amicable relationship" between Altman and Elon Musk "years in the future," while personal expectations involve spending several weeks in the Amazon jungle soon.
- Broader societal and philosophical views suggest AI will act as "gateways" to new realities, with humanity's historical trajectory providing hope, and the expectation that most people will feel gratitude for life even in the face of imminent death.
- Regarding space and existence, it is expected that space travel may be "really difficult," potentially explaining the lack of contact with alien civilizations, though answering such questions via AI remains a high priority.
- Altman expresses no concern about becoming overly cynical but worries about losing a default trusting nature, while anticipating a "non-zero chance" of being surprised by existential risks.