Nvidia’s Jensen Huang: AGI Has Arrived, and 400,000 More GPUs Are Coming
Nvidia chief executive Jensen Huang wrote on X on Sunday that Artificial General Intelligence (AGI) has arrived, crediting OpenAI’s GPT-6 Astra and promising 400,000 more GPUs. The same day, OpenAI’s chief scientist, Jakub Pachocki, called for extreme caution and voluntary slowdowns, emphasizing that no lab has solved alignment and monitoring well enough to keep scaling at maximum speed.
Huang’s Declaration:
"AGI has arrived," Huang declared succinctly on X, congratulating OpenAI on GPT-6 Astra. His post, which had over 10 million views by Monday morning, continued to highlight the use of Nvidia Grace Blackwell chips and the upcoming deployment of 400,000 GPUs.
Context and Reactions:
- Who Claims AGI?: The claim sparked a debate, with OpenAI itself being cautious about the term. While OpenAI’s president, Greg Brockman, thought the company was close, the official account did not use the term "AGI" in its launch posts.
- Benchmarks and Validations: OpenAI’s launch posts focused on benchmarks like FrontierMath Tier 4, ARC-AGI 3, and Terminal-Bench 4.0. The ARC Prize, however, published scores for Astra, indicating a 99.9% accuracy inside OpenAI’s harness and a 62.7% accuracy in the foundation’s standard harness, but did not claim AGI.
- Scientist’s Caution: In a separate essay, Pachocki argued for a time of extreme caution, suggesting voluntary slowdowns as labs and suppliers scale their AI models.
The contrasting views from the supplier and the lab highlight the ongoing debate surrounding AGI’s arrival and the implications for the future of artificial intelligence.