OpenAI just told everyone something wild. Inside its own research team, the AI agents are now putting in more hours than the humans. Roughly three times more.
They call it hitting the automated research intern goal. Back in the fall they set a target: build a system that can take a research task, run with it for days, and come back with real work a person can check. They said they would get there by September. They got there.
The number that stands out is 3.1. That is agent workdays for every one human workday, measured against a normal eight hour shift. A year ago that ratio was flipped. Humans were doing basically everything and the agents were a side tool nobody fully trusted yet.
The typical researcher there is now running more than 600 dollars a day in AI usage, often with several sessions going at once. The heaviest users are burning through thousands a day. That is not someone occasionally asking a chatbot a question. That is a lab where the AI has become a coworker with its own workload.
Here is the part worth sitting with. This is not a demo running in a lab corner. It is the team that builds the next model, using AI to help build the next model, faster than they could alone. OpenAI even said it is aiming for a fully automated AI researcher by 2028.
I get why that number makes people nervous. Machines doing research on machines, moving quicker than a person can double check, is a real thing worth watching closely, not waving off.
But flip it around. Every researcher on that team still opens their laptop, still decides what to try, still says yes or no to what the agent found. The work got faster. The judgment did not get replaced. That is the version of this story worth rooting for, and it is the one happening right now.
——
Follow: @Ali Demi
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277
——
Sources:
https://openai.com/index/research-acceleration-view-inside-openai/
https://www.engadget.com/2251859/openai-says-it-reached-its-goal-of-creating-an-automated-research-intern/
https://www.unite.ai/openai-hits-goal-of-building-an-automated-research-intern/
Repost this. Thanks.
