Automated AI Researcher

An Automated AI Researcher is the term for a target system that, given only broad direction from a human, would design and carry out deep learning and alignment research (the field of making AI act in line with human intent) on its own. It isn't a finished product yet — it's closer to a milestone that OpenAI has reaffirmed it intends to reach by March 2028.

To put it in perspective: if OpenAI's earlier goal, the 'automated research intern,' is like a new hire who completes multi-day tasks when given specific instructions by a senior colleague, the Automated AI Researcher is more like a senior researcher who designs and runs experiments after being given only a broad direction from a team lead. OpenAI has said it already achieved the intern-level goal in this announcement, and has now set this far more autonomous researcher as the next target.

The figure mentioned in the current article — coding agents replacing more than 3x the research labor — is a metric showing an intermediate step toward this goal. In other words, coding agents are already doing a lot of work, but that doesn't mean the Automated AI Researcher has been achieved.

The article states, "They also reaffirmed their plan to build an 'Automated AI Researcher' that would carry out deep learning and alignment research on its own, under human supervision." One point that's easy to get confused about: this term doesn't refer to a system that already exists — it's a goal targeted for March 2028. The current figure showing coding agents replacing 3.1x the researcher labor only reflects one step along the way toward this goal; it doesn't mean the Automated AI Researcher itself has been completed.