AI Agents Do More of the Work in Model Development, but Humans Still Make the Decisions
4 Articles
4 Articles
AI Agents Increase Activity But Not Autonomy — Guardrails To Prevent The Rubber‑stamp Risk » Saipien
AI agents do more of the work in model development, but humans still make the decisions Executive summary Agent activity rose sharply: the median agent-actions-to-human-inputs ratio climbed from 11 to 28.5 over four weeks, the metric counts discrete agent-initiated steps versus explicit human inputs, as reported by the Atria team. Scale without surrender: agents proposed […]
An analysis of 769 task records and agent records of an AI project shows that these systems are increasingly performing work, but people continue to define final goals, methods and decisions. The team also warns of the risk that supervision will end up being reduced to approving hard-to-revise chains of work.
AI agents do more of the work in model development, but humans still make the decisions
A research team analyzed 769 task logs from building its own AI model. AI agents supplied up to 55 percent of method proposals, but humans made more than 85 percent of final decisions. A third of the tasks wouldn't have been attempted without AI. The authors warn that more agent activity doesn't mean more autonomy. The article AI agents do more of the work in model development, but humans still make the decisions appeared first on The Decoder.
A research team evaluated 769 task protocols from the development of its own AI model. AI agents provided up to 55 percent of the method proposals, but more than 85 percent of the final decisions were taken by people. A third of the tasks would not have been addressed without AI. Authors warn against confusing growing agent activity with increasing autonomy. The article AI development with AI agents: AI takes on more work, but hardly makes any d…
Coverage Details
Bias Distribution
- There is no tracked Bias information for the sources covering this story.
Factuality
To view factuality data please Upgrade to Premium





