OpenAI gives ChatGPT new web research skills
Add Axios as your preferred source to
see more of our stories on Google.
Image: OpenAI
ChatGPT is getting a new deep research mode that will let users ask the chatbot to perform "multi-step research on the internet for complex tasks," OpenAI announced Sunday.
Why it matters: It's another step toward fulfillment of the AI industry's promise that AI agents will soon be able to perform human tasks reliably, efficiently and independently.
What they're saying: "Deep research is built for people who do intensive knowledge work in areas like finance, science, policy, and engineering and need thorough, precise, and reliable research," OpenAI said in a blog post.
- "It can be equally useful for discerning shoppers looking for hyper-personalized recommendations on purchases that typically require careful research, like cars, appliances, and furniture."
- "What you get from deep research is a comprehensive, fully cited research paper, essentially something that an analyst or an expert in a field might produce to you," OpenAI chief research officer Mark Chen said on a live-streamed demo of the new tool.
Yes, but: OpenAI also warns that deep research "can sometimes hallucinate facts in responses or make incorrect inferences," though it does so less than other ChatGPT models.
- Deep research also has trouble distinguishing rumors from fact, and "often fail[s] to convey uncertainty accurately."
Deep research will initially be made available to OpenAI's $200/month Pro customers.
- The company says it expects to expand the rollout to include $20/month Plus users within about a month, "if all safety checks continue to meet our release standards."
- Deep research "is currently very compute intensive," OpenAI warns, and it can take from 5 to 30 minutes per task.
The big picture: The release of deep research follows fast on the heels of OpenAI's release of its Operator agent, which can act on behalf of users within a web browser to order products and reserve tickets.
- In the future, OpenAI said in a blog post, "the combination of deep research, which can perform asynchronous online investigation, and Operator, which can take real-world action, will enable ChatGPT to carry out increasingly sophisticated tasks for you."
