Sébastien Bubeck, a recent addition to OpenAI from Microsoft, has made predictions about the future capabilities of AI models. He stated that an AI model will 'for sure' win a gold medal at the International Mathematical Olympiad next year. Bubeck explained that AI model capability can be measured in 'AGI time,' noting that GPT-4 can perform tasks that would take a human seconds or minutes, while another model, referred to as 'o1', can handle tasks measured in AGI hours. He anticipates that within the next year, AI models will achieve capabilities measured in AGI days, and in three years, AGI weeks. This progression reflects a significant compression of human time units, indicating rapid advancements in AI across various domains over the next three to five years.
https://t.co/VUcQyG6eRq Sebastien Bubeck, a new contributor to OpenAI, has an interesting thought; in the course of test-time-compute you can see how smart models are. Previously they thought in AGI-seconds, now in minutes, next year in days. In short: the models think longer.
🚨 Sebastien Bubeck recently joined OpenAI from Microsoft to contribute to advancing AGI He said: AI model capability can be measured in 'AGI time': - GPT-4 can do tasks that would take a human seconds or minutes - o1 can do tasks measured in AGI hours next year, models will… https://t.co/QeANbZop0L
Fully agree with @SebastienBubeck here … AI progress = compressing human time units: now at secs/mins, moving to hrs/days/months/yrs across gradually all domains in next 3-5 yrs. Human time compression = true measure of AI advancement. LFG. https://t.co/1TOTOjj0T1