Is general AI already here? Debate around GPT-6 Astra
In brief
- OpenAI claims GPT-6 Astra may represent AGI by outperforming humans in most economically valuable cognitive tasks.
- Experts challenge this definition, offering stricter criteria or even rejecting the AGI concept entirely.
- GPT-6 Astra’s results vary by benchmark, with top scores relying on a custom software environment.
OpenAI suggests its latest model, GPT-6 Astra, could signal the arrival of artificial general intelligence (AGI), defined as an AI system outperforming humans in most economically valuable cognitive tasks. The company’s definition excludes manual labor like agriculture or construction, though AI-robotics integration may blur this line.
Other experts, including Demis Hassabis (Google DeepMind) and Yann LeCun (Meta), propose alternative definitions, emphasizing the ability to perform « nearly all » human cognitive tasks or even questioning the term AGI itself. Greg Brockman, OpenAI’s president, personally believes AGI has been reached, while Dario Amodei (Anthropic) calls for slowing frontier AI development, citing systemic risks.
GPT-6 Astra’s actual performance remains debated. The model achieved 99.9 % on a specific benchmark (ARC-AGI-3) thanks to an optimized software environment, but independent evaluators downplay this progress. OpenAI also claimed a partial solution to the Navier-Stokes equations, later attributed to an unreleased internal model. Experts stress that current capabilities remain narrow and task-specific, far from general intelligence.
This summary gives you the essentials. The full article, its figures and quotes are on The Conversation : IA.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error