Sources
Loading...
Additional media
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
Bob McGrew, former Chief Research Officer at OpenAI, has articulated a clear path toward achieving artificial general intelligence (AGI), emphasizing the importance of reasoning and test-time compute. In a recent discussion, McGrew noted that overcoming the challenges associated with reasoning, rather than just pre-training, is crucial for AGI development. He highlighted that the focus is now shifting entirely to scaling, which he believes will facilitate significant advancements in AI capabilities. The ongoing progress with OpenAI's Gemini reasoning models has been described as transformative, likening it to the early stages of the GPT-2 era, with improvements being observed on a weekly basis. This development indicates a promising trajectory for future AI systems.
OpenAI former CRO, Bob McGrew — believes the path to AGI is now clear. The problems have been solches and we are now in a pure „scaling regime“. And looking at the pace of how quick o3 was developed, I think he is right. https://t.co/XXqdHAu8Hl
OpenAI former CRO, Bob McGrew — believes the path to AGI is now clear he explains, - the biggest challenge in building AGI was reasoning, not just pre-training - now that this has been solved, the focus is entirely on scaling, which marks a clear path for AGI https://t.co/Kcl8v6Cq20
"The progress with our Gemini reasoning models is actually wild, we are in the GPT-2 era of scaling reasoning!" I always think of the words of Noam Brown (OpenAI), who said that the pace of development and progress is continuing (scaling). From o1 to o3 three months. From o3 to… https://t.co/PaNLnLDMwW