A research team has conducted an analysis of 769 task logs from the development of an AI model, revealing that AI agents can supply up to 55 percent of proposals for methods. Despite this substantial contribution, humans are responsible for more than 85 percent of final decisions made during the model development process.
The findings underscore a critical point: increased activity from AI agents does not equate to greater autonomy. In fact, the study notes that one-third of the tasks analyzed would not have been attempted without the involvement of AI. This highlights the ongoing necessity of human oversight in AI development, even as agents become more integrated into the process.
Compiled automatically by the Tech AI Newsdesk from public AI-news sources and summarised in our own words.
