Isla Fulford, a researcher from OpenAI, had a feeling Deep research He would be a hit before it was released.
Fulford has helped build an artificial intelligence agent, which autonomously explores the network, deciding about yourself, what links to click, what to read and what to combine in a thorough report. Opeli for the first time made deep examinations available internally; Whenever it fell, Fulford says that she was flooded with questions from her colleagues willing to recover him. “The number of people who gave me very much excited us,” says Fullord.
Since the audience, on February 2, deep research turned out to be a hit of many users outside the company.
“Deep research has written 6 reports so far”, Patrick Collison, CEO of Stripe Posted to X A few days after the product is released. “It’s really perfect. Congratulations to people behind it.”
“Deep research is an AI product, which really has a significant part of the community of creating politics in DC to start feeling agi” wrote Dean Ball, a member of George Mason University, who specializes in AI’s policy.
Deep research is available as part of the ChatGPT Pro plan, which costs $ 200 per month. This takes a question, such as “write to me a report on the health insurance industry in Massachusetts” or “Tell me about the WIRED covering regarding the Department of Government Efficiency”, and then develops a plan, searching for the relevant websites, combing their content and deciding what links click and what information deserves further research. After examining tens of minutes for time, it synthesizes its arrangements into a detailed report, which may include quotes, data and charts.
Many tools currently marked as AI agents are basically chatbots combined with straightforward programs without much sophistication. The deep research model itself undergoes artificial reasoning before developing the plan and going forward with each step. The model contains details of this reason for research in the side window.
“Sometimes it’s like” I have to go back, it doesn’t seem so promising, “says Josh Tobin, another Openai researcher involved in building deep research. “It’s nice to read some of these trajectories to understand how the model thinks.”
Openai apparently perceives deep research as a tool that can undertake more office work. “This is a thing we can scale,” says Tobin, adding that the agent can be trained to complete specific white work. An agent with access to internal data of the company can, for example, quickly prepare a report or presentation. Tobin claims that a longer goal is to “build an agent who is not only good in building reports by searching the network, but is good in many other types of tasks.”
Since the deep research was trained to analyze and summarize the text written by man, Tobin says that his team was surprised, seeing that many people use him to generate code. “This is an compelling thread to pull,” he says. “We are not entirely sure what to do about it.”