Do you really want the US to "win" AI?

TLDR

• George Hotz argues that US AI dominance is only worth wanting if AI reaches ordinary people as hard possession, not a revokable API privilege.

Key Takeaways

• Hotz says he self-identifies as someone who should embrace the current AI moment but rejects the neofeudalist society it is producing.
• In the Altman vs. Musk lawsuit, Hotz sides with Altman: Altman is a product guy who builds things people use; Musk has released nothing seriously open source.
• Hotz calls out Anthropic for repeating the same fear-based marketing playbook they ran at OpenAI in 2019 with GPT-2 XL, noting the founders are literally the same people.
• SpaceX diluting its Mars mission by buying AI bubble assets frustrates Hotz; he sees hard colonization projects and soft Earth society as needing different values.
• The test Hotz proposes: pay attention to who has actually released AI to the world versus who controls it through an API they can revoke.

Why It Matters

• Hotz draws a concrete line between open AI (Linux, ffmpeg model) and controlled API AI, arguing only the former serves normal people.
• American workers have watched tech companies extract value for a decade; Hotz says giving those same companies more AI power makes extraction worse, not better.
• The argument reframes the “US winning AI” debate away from geopolitics toward who inside the US actually holds the capability.

the singularity is nearer · 2026-04-22 · Read the original