OpenAI has introduced a new AI tool, Deep Research. It is an innovative agent designed to facilitate deep and complex research on the ChatGPT platform.
According to the OpenAI blog, this feature is designed for professionals working in fields such as finance, science, politics, and engineering who need deep, accurate, and reliable information. At the same time, it can also help users who need to thoroughly research large purchases such as cars, appliances, or furniture.
Deep Research, not superficial answers
Deep Research is designed for situations that require in-depth analysis of data from various websites and sources, rather than a quick answer or summary. The new tool is available today for ChatGPT Pro users for up to 100 queries. It is planned to be expanded to Plus and Team users next month, and then to Enterprise customers.
Usage process and features
Users can select the “deep research” option, enter questions and attach files or spreadsheets if necessary. For now, this only works on the web version, but integration is expected for the mobile app later this month. Depending on the complexity of the question, it can take from 5 to 30 minutes to get results, and the user will be notified when the search is complete.
In the first phase, the answers will be text-only, but OpenAI has announced that it plans to include images, graphs and other analytical data in the near future. It will also be possible to connect to paid databases and internal resources.
Accuracy issue
AI technology is not perfect and is prone to errors. Therefore, OpenAI emphasized that all answers are provided with accurate sources and links, making it easy for users to verify the information. Meanwhile, a new AI model called o3 has been optimized specifically for the service. It has learned to use web search and Python tools to solve real-world problems.
OpenAI has tested this model and reported that it achieved an accuracy of 26.6% on the Humanity’s Last Exam test, which consists of complex questions. While this number may seem small, it is considered a high result because the test is much more difficult than other standards.
Deep Research simplifies tasks that require deep analysis in many areas. However, OpenAI admits that it sometimes makes mistakes. More advanced versions are expected in the future, with access to more data. Now, the deep learning process can be carried out on the basis of more systematic and detailed sources.
