This definitely sounds like something I've heard before.
There is little doubt that OpenAI has developed a fresh sense of urgency. On Sunday evening, just two days after the o3-mini was made available to the public, the business made an unexpected revelation about Deep Research. ChatGPT's new function enables it to locate, evaluate, and combine information from hundreds of websites and online sources in order to generate reports that are comparable to those produced by a research analyst.
In addition to the standard text inquiries, users can upload files, such as PDFs and spreadsheets, when they prompt ChatGPT in this manner. After that, the chatbot will take "anywhere from 5 to 30 minutes" to put together a response, and a side panel will keep track of the agent's progress and citations while it is working. OpenAI states that the new capability "accomplishes in tens of minutes what would take a human many hours."
Mark Chen, the chief research officer of OpenAI, stated during the company's reveal livestream, "Our ultimate aspiration is a model that can uncover and discover new knowledge for itself." "It is a fundamental part of our [artificial general intelligence] plan."
According to OpenAI, ChatGPT has several limitations. For example, it can sometimes make false inferences or hallucinate facts when undertaking Deep Research. However, it does so "at a notably lower" rate than other contemporary models. Furthermore, the agent may occasionally have difficulty distinguishing between credible facts and rumors. Some users may also notice that there are formatting problems. The business states, "We anticipate that all of these problems will improve quickly with increased usage and time."
If any of this seems familiar to you, it is because Google's Advanced suite has its own Deep Research component. This feature not only has the same name, but it also has a similar range of capabilities. The $20 per month One AI Premium package from Google provides access to Gemini Advanced, which is one of the major differences between the two. On the other hand, if you want to use OpenAI's version of Deep Research today, you will have to pay $200 a month for a ChatGPT Pro membership.
The company explains, "Deep research in ChatGPT is currently very compute intensive," and adds that Pro customers would be limited to 100 inquiries each month. "The longer it takes to research a query, the more inference compute is needed."
According to OpenAI, the company is developing a version of Deep Research that is powered by a smaller and more affordable model. As a result, the corporation will be able to provide "significantly higher rate limits." In the meanwhile, OpenAI is hoping to make the tool available to Plus users "in about a month," after conducting a round of safety testing. European users will have to wait to try out the tool for themselves, just like they have had to do with most of the company's other recent releases. Deep Research is not yet available to anyone in the UK, Switzerland, or the broader European economic zone.
