Advertisment

OpenAI launches deep research agent for ChatGPT

The deep research agent for ChatGPT enables in-depth analysis with citations, using an o3 model optimised for web browsing and data interpretation.

author-image
Social Samosa
New Update
BVB

OpenAI has introduced a new AI-powered agent designed to assist users in conducting comprehensive research through its ChatGPT platform. Called 'Deep Research,' the tool is aimed at individuals in fields such as finance, science, policy, and engineering who require precise and reliable information. It is also useful for consumers making high-stakes purchases, such as cars, appliances, and furniture.

Image :OpenAI 

Available to ChatGPT Pro users from today, Deep Research allows users to request more thorough responses, involving information from multiple websites and sources. The feature is designed for situations where a quick answer won’t suffice, with results taking anywhere from five to 30 minutes. At present, the tool is accessible via web browsers only, with mobile and desktop app integration expected later this month. It will initially allow 100 queries per month, with plans to expand availability and increase query limits for paid users soon.

Deep Research outputs are currently text-based, but OpenAI plans to include images, data visualisations, and other analytics in future updates. The tool will also support more specialised data sources, including subscription-based resources. To ensure accuracy, each response will be fully documented, with citations and a summary of the reasoning behind the information provided.

Image: OpenAI

The deep research tool is powered by OpenAI's o3 'reasoning' AI model, optimised for web browsing and data analysis. Trained using reinforcement learning on real-world tasks, the model is designed to search, interpret, and analyse vast amounts of information, including text, images, and PDFs. It can also browse user-uploaded files and generate graphs. However, OpenAI acknowledges that the AI is not flawless, often making mistakes, formatting errors, and struggling to differentiate between credible information and rumours.

OpenAI tested the deep research tool using a challenging set of questions from 'Humanity’s Last Exam,' which includes over 3,000 expert-level academic questions. The o3 model achieved an accuracy of 26.6%, outperforming competitors such as Gemini Thinking and Grok-2.

Despite these advancements, OpenAI cautions that deep research still has limitations, particularly in cases where uncertainty is involved. It may not always identify when it is unsure about certain information and can sometimes make incorrect inferences.

The launch of Deep Research follows a similar announcement by Google, which introduced a comparable feature just two months ago. However, OpenAI’s deep research tool aims to provide users with a more robust, well-researched output, equipped with citations to back up its claims. Whether users will adopt this more meticulous approach to AI-generated information remains to be seen.

Chat GPT Open AI Deep Research agent