The advanced research tool of ChatGPT can generate reports from hundreds of online sources.
After the release of o3-mini, OpenAI announced Deep Research, a new agentive tool for ChatGPT, on Sunday.
OpenAI has recently ramped up its efforts, evidenced by the launch of a new feature called Deep Research, unexpectedly unveiled two days after the announcement of o3-mini. This tool allows ChatGPT to search, analyze, and synthesize information from hundreds of websites and online sources to generate reports with a quality comparable to that of a research analyst.
Users can interact not only through standard text questions but also have the option to upload files, such as PDFs and spreadsheets. ChatGPT takes between five and thirty minutes to compile a response, and a sidebar shows the progress of the analysis and the citations used. According to OpenAI, this new feature can accomplish in a short time what would take a human many hours.
Mark Chen, OpenAI's research director, commented in a live broadcast that the aim is to develop a model that can discover and create new knowledge on its own, which is fundamental to their roadmap in artificial general intelligence.
However, some limitations have been noted. OpenAI warns that ChatGPT may occasionally generate incorrect information or make erroneous inferences during deep research, although with a "notably lower" error rate than other current models. It has also been found that the system sometimes struggles to distinguish between authoritative information and rumors, and it may present formatting errors. The company is confident that these issues will improve over time and with usage.
It is important to highlight that the Deep Research feature is not exclusive to OpenAI; Google also offers a similar feature in its Advanced suite. However, the difference lies in that access to Gemini Advanced is available through a $20 monthly premium plan, whereas using OpenAI's Deep Research feature requires a $200 monthly ChatGPT Pro plan.
The company also mentioned that Deep Research is currently resource-intensive, which will limit Pro users to 100 queries per month. To optimize the resource, OpenAI is developing a version of Deep Research that would operate with a smaller and more economical model, allowing for higher query limits. In the near future, the intention is for Plus users to also gain access to this tool once a safety testing round has been completed. However, users in Europe will still have to wait, as Deep Research is currently not available in the UK, Switzerland, and the European Economic Area.
