Tag: multistep

Why It Matters Cutting corners: OpenAI has introduced a new tool called Deep Research within its ChatGPT platform. This tool utilizes reasoning to synthesize vast amounts of information, complete multi-step research tasks, and generate detailed reports. The launch of Deep Research comes at a time when there is increasing competition, notably from China's DeepSeek, which recently released a free chatbot app based on its R1 model, causing a stir in the global AI community. Deep Research is designed to perform tasks on behalf of users, creating comprehensive reports by consolidating data from online sources, much like a research analyst would. It is powered by a version of OpenAI's upcoming o3 model, capable of searching, interpreting, and analyzing extensive text, images, and PDFs available on the internet. According to OpenAI, Deep Research can accomplish tasks in just "tens of minutes" that would typically take humans many hours to complete. The company envisions this tool as a significant step towards its overarching goal of developing artificial general intelligence (AGI), a concept that some experts believe could lead to the production of groundbreaking scientific research. The new tool, accessible as a button within the ChatGPT app, targets professionals in knowledge-based sectors such as science, engineering, finance, and policy. These fields demand accurate and reliable research, which Deep Research aims to deliver. Additionally, the tool can assist general users by providing tailored recommendations for purchases like cars, appliances, and furniture, tasks that typically involve extensive manual research. To utilize Deep Research, users need to select it in the ChatGPT message composer and input their query. OpenAI asserts that the tool can provide precise insights on a variety of topics, ranging from streaming platform analysis to personalized reports on the best commuter bikes. Deep Research typically takes between five and 30 minutes to conduct complex web research tasks. Upon completion, users receive a notification signaling the availability of the generated report. OpenAI is also working on integrating embedded images, data visualizations, and other analytical outputs into reports to enhance clarity and context. This announcement follows OpenAI's recent introduction of another AI agent called Operator, tailored for e-commerce users. Operator can facilitate tasks such as booking flight tickets, making restaurant reservations, and online shopping using a photo of a shopping list. Currently, Operator is exclusive to ChatGPT Pro users in the U.S., with plans for broader availability in the future. OpenAI's latest cost-efficient reasoning model, o3-mini, has also been unveiled to counter DeepSeek's advancements. o3-mini is now accessible through ChatGPT and via API, offering a specialized model for STEM fields while keeping operational costs low.

Tag: multistep

OpenAI debuts “Deep Research” model to tackle multi-step research AI tasks

Quick Links

Resources

Support