Sounds helpful.Agent oo3OpenAI has unveiled a new research tool that can pull information from the web for you and summarize it into detailed reports supposedly "at the level of a research analyst." Naturally, major caveats to follow.Called "deep research," yes, without capitalization the tool is a type of AI agent, an industry term used to describe models that can perform tasks on your behalf. OpenAI released its first AI agent, Operator with the "O" capitalized only last month, showcasing its ability to shop for groceries and make reservations online.Catering towards higher-brow ambitions, OpenAI has built deep research for "intensive knowledge work" in fields like finance and science. But for those who like the idea of a bot doing their shopping, OpenAI says deep research, which is powered by an upcoming version of the company's o3 model, can help in that department by giving them "hyper-personalized" recommendations on big purchases like cars and appliances."It accomplishes in tens of minutes what would take a human many hours," OpenAI claimed, by leveraging its reasoning capabilities to peruse text, images, and PDFs on the internet.GoogleGPTDeep research is available only to those who are subscribed to the $200-per-month ChatGPT Pro plan. In addition to giving text responses like the chatbot normally does, the tool also features an activity sidebar that shows a step-by-step summary of its process in real time. Right now, its final reports are strictly text-based, but OpenAI says that images and data visualizations will be added in the coming weeks.Given that it's actually opening and looking through potentially enormous amounts of web content, responses can take anywhere from five to 30 minutes to produce, OpenAI said. But once you give it its instructions and answer some preliminary questions, you can step away, and the bot handles the rest.Fact CheckThere's no guaranteeing the rigor of the research, however. Like all large language models, deep research can "sometimes" hallucinate or make up facts, OpenAI admitted, though the company claims the tool does this at a lower rate than its existing models. The AI agent also struggles with separating "authoritative information from rumors," and often fails to convey uncertainty. In other words: it has a bad habit of passing fiction off as fact.Impressive as the tech may be, those are huge drawbacks. We're not talking about a plain ol' chatbot anymore. We're talking about something explicitly geared towards serious knowledge gathering, even in scientific fields. Less leeway is permissible here.Deep research may be able to gather loads of information in a matter of minutes, but how much of it can be trusted? And how much time would you spend double-checking the accuracy of its synthesized report, and by extension the sources it cites? These are all questions worth asking as OpenAI rushes another AI out the door, after having a fire lit under it by DeepSeek.Share This Article