Introduction to Deep Research

In the video, Mark from OpenAI introduces “Deep Research,” a new capability that allows AI agents to perform multi-step research on the internet, providing comprehensive, fully cited reports similar to human analysts. The team demonstrates its versatility through various applications, highlighting its ability to handle complex queries and streamline the research process, all powered by a fine-tuned version of the upcoming O3 reasoning model.

In the video, Mark, the head of research at OpenAI, introduces a new capability called “Deep Research,” which aims to enhance the functionality of AI agents in knowledge work. Joined by team members Issa, Josh, and Neil, they discuss the significance of agents in transforming both enterprise processes and consumer experiences. The introduction of Deep Research follows the launch of the O1 model, which was the first in a series of reasoning models designed to think deeply before providing answers. However, a limitation of these models was their lack of access to tools, particularly the internet, which Deep Research aims to address.

Deep Research is designed to perform multi-step research on the internet, allowing it to discover, synthesize, and reason about content while adapting its approach as it gathers more information. Unlike traditional models that return answers quickly, Deep Research can take anywhere from five to thirty minutes to provide comprehensive, fully cited reports akin to those produced by human analysts. This extended processing time is seen as beneficial, enabling the model to undertake more complex and autonomous tasks, which aligns with OpenAI’s roadmap towards artificial general intelligence (AGI).

The video showcases how Deep Research can be utilized for various applications, including market research and personal projects. Neil demonstrates the functionality by querying Deep Research for information on iOS and Android adoption rates, emphasizing the model’s ability to ask clarifying questions to ensure accurate results. This interaction highlights the model’s capability to handle complex queries that would typically require significant manual effort, thus streamlining the research process for users.

Josh further illustrates the versatility of Deep Research by using it to find recommendations for purchasing skis while in Japan. He emphasizes the importance of specifying requirements to receive tailored results. The model’s ability to browse the internet, analyze multiple sources, and present findings in a well-structured report format is showcased, demonstrating its potential for both professional and personal use cases.

Finally, the team discusses the underlying technology of Deep Research, which is powered by a fine-tuned version of the upcoming O3 reasoning model. This model has been trained using reinforcement learning to effectively plan and execute complex tasks. The video concludes with Mark reiterating the significance of Deep Research in advancing AI capabilities and its potential applications in various domains, while also inviting users to share their experiences with the new feature.