@心悦Joy good article sharing
Post
--
心悦Joy
·
--
Recently, OpenAI’s official blog published an article titled “Research acceleration: The view inside OpenAI,” and for the first time publicly disclosed progress on AI “recursive self-improvement” (RSI) in the form of internal data. The article notes that the team has already achieved the goal set last fall: building an “automated research intern” by this September. An “automated research intern” refers to a system capable of carrying out clearly defined research tasks under human guidance, including tasks that experienced researchers would need several days to complete. Looking ahead, OpenAI expects to achieve a full “automated AI researcher” by March 2028, enabling it to participate in deep learning and alignment research and further improve the system through iteration. This means the vision of “AI training AI” is accelerating.
Related data also indirectly supports this point. According to OpenAI’s statistics, at the beginning of this year, the median use of Agents by OpenAI’s internal researchers was still relatively low. By mid-August, those researchers had integrated Agents into their daily work. Based on public API pricing, researchers at median usage levels were already consuming token value worth more than $600 per day; among the organization’s top 10% heavy users, daily token consumption exceeded $7,000. Image source: OpenAI
In addition, OpenAI noted that as automation advances, the research tasks least susceptible to automation will instead take up more of researchers’ time and become the future bottleneck in AI R&D; once other bottlenecks are eased, the importance of compute will become even more pronounced. Shortly after the article was published, OpenAI team member Kevin Liu reposted it on X and said: “RSI is very likely to be the core driver of the next leap in AI capabilities over the coming years, but this technology currently exists only in a few leading AI labs.”
He further called for: “The importance of information transparency has never been greater. It can help the public fully discuss: at what pace should we advance model development, and whether it needs to be constrained. I recommend that other AI companies take the same approach to public disclosure.” In fact, OpenAI is by no means the only AI company betting on RSI.
Related data also indirectly supports this point. According to OpenAI’s statistics, at the beginning of this year, the median use of Agents by OpenAI’s internal researchers was still relatively low. By mid-August, those researchers had integrated Agents into their daily work. Based on public API pricing, researchers at median usage levels were already consuming token value worth more than $600 per day; among the organization’s top 10% heavy users, daily token consumption exceeded $7,000. Image source: OpenAI
In addition, OpenAI noted that as automation advances, the research tasks least susceptible to automation will instead take up more of researchers’ time and become the future bottleneck in AI R&D; once other bottlenecks are eased, the importance of compute will become even more pronounced. Shortly after the article was published, OpenAI team member Kevin Liu reposted it on X and said: “RSI is very likely to be the core driver of the next leap in AI capabilities over the coming years, but this technology currently exists only in a few leading AI labs.”
He further called for: “The importance of information transparency has never been greater. It can help the public fully discuss: at what pace should we advance model development, and whether it needs to be constrained. I recommend that other AI companies take the same approach to public disclosure.” In fact, OpenAI is by no means the only AI company betting on RSI.
Disclaimer: Includes third-party opinions. No advice. Binance AI may be used without guarantee. See T&Cs.