OpenAI releases new ChatGPT feature for in-depth research
Share
OpenAI has introduced a new feature for ChatGPT called Deep Research. It allows artificial intelligence to work autonomously, planning and executing multi-stage research.
This was reported by The Verge.
What the new Deep Research feature can do
The tool is able to search for the necessary data, analyze it, and adapt the search strategy in real time.
Unlike a standard text generator, Deep Research displays progress in a sidebar along with sources and summary analysis. Users can ask questions in text form and attach images, PDFs, or tables for context. The answer can be generated in 5 to 30 minutes. In the future, OpenAI plans to add embedded graphics and images to the research results.
Despite the new possibilities, OpenAI warns that Deep Research is not perfect:
- sometimes it can invent facts (so-called "hallucinations");
- may confuse authoritative information with rumors;
- does not always correctly assess the level of confidence in its conclusions.
The development of AI-based agents is becoming an important area for OpenAI and other companies. Deep Research is positioned as a research analyst level tool. The demonstration video shows a query about changes in retail over the past three years, to which the artificial intelligence responds with tables and lists.
The new tool comes shortly after the launch of the OpenAI Operator, which uses a web browser to perform tasks. Google is also testing a similar prototype, Project Mariner, but it is not yet available for public use. Instead, Deep Research has already been launched in the version for Pro subscribers.
ChatGPT Pro users get up to 100 queries per month. OpenAI also promises "limited access" for Plus, Team, and Enterprise subscribers. Due to the high computing power required for deep research, the company plans to raise the limits for all paid users in the future.
As a reminder, the U.S. will further restrict the export of artificial intelligence chips and technologies, dividing the world to keep advanced computing power in the United States and its allies.
Previously, OpenAI accused DeepSeek of illegally using data to train its AI models. Nevertheless, the American company has repeatedly violated copyrights itself.
