ทันเอไอ

"AI 2040" plan to delay Superintelligence until 2040 criticized by George Hotz as mere "cult of intelligence"

Renowned hacker George Hotz criticizes the "AI 2040" proposal from a former OpenAI team, which aims to slow AI development and open-source all research, calling it a concept that overlooks physical realities.

This week's hot topic in the tech community is a critical article by George Hotz (geohot), hacker and founder of Comma.ai, concerning the "AI 2040: Plan A" proposal. This proposal, presented by a group of researchers who previously worked at OpenAI, offers a positive vision for the future of artificial intelligence.

The "AI 2040" proposal extends the "AI 2027" report written by the same team. In the earlier version, they predicted a worrying future where AI could lead to human extinction or totalitarian power centralization. However, "Plan A" proposes an idealistic solution: delaying superintelligence development until 2040, open-sourcing all AI research to decentralize technology, and creating a system for joint monitoring and curbing overly rapid development. The authors include prominent AI safety figures like Daniel Kokotajlo, Thomas Larsen, Eli Lifland, and Romeo Dean.

However, George Hotz countered in his blog, calling this concept part of a "cult of intelligence." He argues that intelligence is not the ultimate goal but merely a current bottleneck. Hotz points out that global domination cannot be achieved solely with "tokens" (basic data units processed by AI models) but requires control over real-world resources and supply chains.

Furthermore, Hotz criticized the AI 2040 report's illustrations of large floating data centers, labeling them as vaporware (software or hardware announced but never materialized). While easy to visualize, such projects would face immense engineering and logistical challenges. This sparked a widespread debate on Hacker News, with over 103 comments and 75 upvotes, reflecting extremely divergent views on future AI development.