OpenAI has unveiled an innovative research tool known as “deep research,” which aims to streamline the process of gathering information from the web and generating comprehensive reports similar to those created by research analysts. This advanced tool, powered by the o3 model, is particularly tailored for demanding knowledge-driven professions, including finance and scientific research.

In contrast to its earlier version, Operator, which primarily assisted with everyday tasks like grocery shopping and booking reservations, deep research is geared towards more complex and nuanced projects. It can provide tailored suggestions for significant investments such as vehicles and home appliances, asserting that it can achieve results in mere minutes, tasks that might traditionally consume hours of manual effort.

This feature is exclusively available to users subscribed to the $200-per-month ChatGPT Pro plan. Deep research functions by meticulously scanning text, images, and PDFs found online. While the response time can vary from five to thirty minutes, users have the advantage of monitoring the tool’s progress through a dedicated activity sidebar. Currently, the reports generated are text-only, but OpenAI has plans to enhance this by integrating images and data visualizations in the near future.

Despite its remarkable features, deep research is not without its drawbacks. OpenAI has recognized that, similar to other large language models, this tool might sometimes produce inaccurate information and may struggle to differentiate between trustworthy sources and mere speculation. Such inconsistencies raise valid concerns regarding the dependability of the reports it generates, particularly in scientific applications.

While deep research can significantly speed up the process of information collection, it is essential for users to remain vigilant and validate the accuracy of its outputs. As OpenAI continues to advance its AI technologies, the importance of careful scrutiny and critical analysis becomes increasingly vital to uphold the reliability of the information provided.