Updated
Updated · The Verge · Aug 27
Nvidia CEO Says AGI Is Achieved for Many Tasks, Calls Milestone Senseless
Updated
Updated · The Verge · Aug 27

Nvidia CEO Says AGI Is Achieved for Many Tasks, Calls Milestone Senseless

3 articles · Updated · The Verge · Aug 27

Summary

  • On Nvidia’s earnings call, Jensen Huang said the company has already achieved AGI “for many tasks,” then argued the milestone itself is “senseless.”
  • Huang offered no fixed benchmark, instead pointing to AI agents that can learn skills recursively and stressing usefulness over labels—systems doing productive work and generating profitable tokens.
  • He has made the claim before: on Lex Fridman’s podcast in March, Huang said “we’ve achieved AGI,” while also conceding the odds of 100,000 agents building Nvidia are “zero percent.”
  • The remark lands in a field with no agreed AGI definition, where OpenAI, Anthropic, Meta, Microsoft, Amazon and Google DeepMind all use competing terms or thresholds.

Insights

If defining AGI is senseless, what concrete metric will actually determine the future of global AI hardware demand?
Can the tech industry safely regulate advanced AI if top executives dismiss its very definition as a meaningless buzzword?
Are profitable tokens enough to prove intelligence, or is the industry simply masking its inability to achieve open-ended reasoning?