- Arthur Mensch said AGI, the industry's holy grail, is just "a marketing move."
- The Mistral CEO argued at London Tech Week that a different metric to judge an AI's progress.
- Mensch said it should be measured by how long agents can reliably run real tasks.
Artificial general intelligence, or AGI, is the AI industry's holy grail — but the CEO of a major AI startup has dismissed it as a "marketing move."
AGI is broadly used to describe a system that could perform any task that a human can.
Speaking during a fireside chat at London Tech Week on Tuesday, Arthur Mensch, CEO of French startup Mistral, said that AGI is "not super well defined."
"It's a moving target by definition," Mensch said. "We move from AI to AGI at some point, but that's a marketing move."
OpenAI's Sam Altman has said that AGI could come as soon as this year, while Google DeepMind's Demis Hassabis predicts it'll arrive by 2030.
The Mistral CEO, known for his AGI scepticism, said that progress should instead be measured by looking at "the length of the agent execution," which means how long an AI agent can perform tasks before it runs into problems.
"The number of tasks that you can do within an enterprise, or as a user using generative AI systems, is definitely growing, and it's going to continue to grow," he said.
Mensch cofounded Mistral in 2023 with former DeepMind and Meta researchers. It has quickly become one of Europe's most prominent AI companies, often seen as the continent's answer to OpenAI.
The startup focuses on open-source AI models and has been scooping up talent from rival Meta, Business Insider previously reported.
Read next
Where Big Tech secrets go public — unfiltered in your inbox weekly.
