OpenAi announces deep research days after the earthquake of Deepseek

Deepseek, the Chinese AI company, caught the entire AI market through surpriseAnd no one more than OpenAi. The Chatgpt-like model R1 received raging reviews with a fraction of the development costs. President Trump called it one “Wake up” Call for the entire American technical industry. Now, just a few days later, OpenAi has fallen back: yesterday the company released a new agentic capacity: Deeply investigation.

This new model, accessible via a button on the chatgpt app and website (only available for pro users, for now), can demonstrably versatile online research, analyze, analyze, synthesize and interpret huge amounts of varying data types (text, graphics, graphics, Graphs, PDFs and more) in 5 to 30 minutes, compared to hours or days of work by a person. Developers say that it has overcome the gap of his previous reasoning model, adapted to technical domains, but which is lacking in order to accept contextual challenges from the real world.

The name choice of OpenAi, deep research, apart from playing deep chat, intentionally or not, is provocative. Are these Gen-AI reports real research? And more specifically, can they replace scientific research? “The possibility of synthesizing knowledge is a condition for creating new knowledge,” Openai wrote in their introductory post. “For this reason, deep research marks an important step in the direction of our broader goal of developing Agi, which we have long been able to produce new scientific research.”

Can agentic AI really produce new insights, insights and creative discoveries? We, researchers, regard our work as more than synthesis and rigorous repetitive validations; It often comes down to that moment of inspiration, a spark of illusive creativity that keeps us going. Not so different, we often feel, of the spark of artistic creativity. OpenAi has already challenged the latter with Sora. A year later, almost to the day, it challenges our concept of scientific creativity. Is reasoning and consolidating huge amounts of methods and existing hypotheses sufficient to make the leap to a discovery? The application of known methods to data is likely to be accelerated and faster steps will leave a previously planned route. Whether agent AI models can really produce new methods and challenge existing knowledge, it still needs to be proven.

Honest research always recognizes its own limitations. The new Chatgpt model is still susceptible to hallucinations and reconstruction. It is still, as the company explains, not transferring uncertainty about his own findings. What does this mean when such models will be included Action those? Openai was rolled out last month OperatorA model that can perform real Real-World tasks for users. For the time being, booking a table in a restaurant or ordering food from a shopping list does not require enormous verification. But think of the day that research can be taken into action immediately. A great concern and enormous potential.