OpenAI claims the software represents a major step towards its overarching objective of creating synthetic basic intelligence (AGI) that matches (or surpasses) human efficiency. It says that what takes the software “tens of minutes” would take a human many hours.
In response to a single question, akin to “Draw me up a aggressive evaluation between streaming platforms,” Deep Analysis will search the online, analyze the data it encounters, and compile an in depth report that cites its sources. It’s additionally in a position to attract from recordsdata uploaded by customers.
OpenAI developed Deep Analysis utilizing the identical “chain of thought” reinforcement-learning strategies it used to create its o1 multistep reasoning mannequin. However whereas o1 was designed to focus totally on arithmetic, coding, or different STEM-based duties, Deep Analysis can deal with a far broader vary of topics. It might additionally alter its responses in response to new information it comes throughout in the middle of its analysis.
This doesn’t imply that Deep Analysis is immune from the pitfalls that befall different AI fashions. OpenAI says the agent can generally hallucinate details and current its customers with incorrect info, albeit at a “notably” decrease charge than ChatGPT. And since every query might take between 5 and half-hour for Deep Analysis to reply, it’s very compute intensive—the longer it takes to analysis a question, the extra computing energy required.
Regardless of that, Deep Analysis is now accessible at no additional price to subscribers to OpenAI’s paid Professional tier and can quickly roll out to its Plus, Crew, and Enterprise customers.