TapTechNews June 12 - According to TradingView, Google software engineer François Chollet expressed his concerns about the current state of AGI (Artificial General Intelligence) research when chatting with host Dwarkesh Patel on a podcast.
Chollet said that all the most advanced research results were publicly shared and published a few years ago, but that is no longer the case now. He attributes this change to the influence of OpenAI, accusing them of causing the 'complete closing down of frontier research publishing'.
Chollet criticized OpenAI for triggering the hype around large language models, believing that it has diverted resources and attention, causing other potential AGI research areas to be overlooked. He said that OpenAI basically set back progress towards AGI by quite a few years, probably like five to ten years.
Chollet recalled the early days of AI research, stating that although there were fewer people involved then, due to exploring various different directions, the progress seemed to be faster. He laments the current state of the field as everyone now seems to be doing variations of the same thing.
Chollet and Mike Knoop created a competition called ARC-AGI in 2019 with a $1 million bonus (TapTechNews note: currently about 7.264 million Chinese yuan). This competition measures the ability of AGI to acquire new skills and efficiently solve novel and open problems.
Knoop said that although 300 teams tried ARC-AGI last year, the score of the most advanced (SOTA) only increased from the initial 20% to 34%, while the human score is between 85% and 100%.
Chollet said that the goal of setting up this award is to increase the number of researchers focusing on cutting-edge AGI research, instead of fiddling with large language models.