Openai will do his Deep research model Available for users at the free level of ChatGPT soon, according to X post by Director General Sam Altman. Deep research is the family name for large language models that use a “test-time” computer method to produce detailed and detailed research reports. These reports can last from 30 minutes to multiple hours to produce, and technologists have said that deep research could replace people in jobs such as business analysis or finance. Of course it is important to take these claims with grain of salt.
Altman also said that Openai would consolidate his numerous AI models. Many people have complained that everyday users do not understand the difference between its various models and what they offer. “We hate the model selector as much as you and want to return to magical unified intelligence, “Altman said.” We will subsequently send GPT-4.5, the model we called Orion inside, as our last non-chain-thought model. After that, a top goal for us is to unify O-series models and GPT series models by creating systems that can use all our tools, know when to think for a long time or not, and is generally useful for a very wide range of tasks. ”
At some point in the near future, users on the free level will gain “unlimited” access to a new model, which includes an O3-built, although Altman says that paid users will receive a higher level of intelligence, whatever it means.
Currently deep research is only available to ChatGPT users who pay for its prize, a $ 200-month level. It has always been expected that Openai to provide its models from bleeding edge to more users like it scales its server infrastructureand Altman confirmed so much in another recent X post. In his typical way to write everything lowercase as a Gen Z girl, Altman wrote, “I think we will initially offer 10 uses a month for Chatgpt Plus and 2 a month in the free level, with the intention to scale these with time.”
Large language models like ChatGPT are, essentially, probable machines, or a really powerful car set. They ingest billions of pieces of human written text and are able to generate new texts with the versatility of something written by a person. They actually don’t “think” as a person does, and companies like Openai had to find new techniques to improve their accuracy and skills. Test-time thinking forces the language model fact-control itself before answering a question. For example, if a user asks the model how much would cost to replace each Uber in the United States with Waymos, the model would break the survey in multiple related questions – how many Ubers are today on the road, how much every Waymo vehicle costs, and so on .
The AI industry threw spaghetti at the wall some time ago, trying to find the unlock that will make language models really capable of replacing people in various tasks. Deep research is the latest such test; Another is “agents” who can Take control of the user’s computer To actually perform tasks, such as reservation flight. Agents could be particularly useful for demographics such as the elderly, who may not be technological and could ask to help them navigate a computer. Early users of agents like Openai’s operator say that they are still very slow and wrong.
As impressive as deep research could produce long research reports, the accuracy of AI -language models remains a problem. While the Super Bowl, Google issued business advertising Incorrect information. X users have noticed if you ask ChatGpt, what players are in various NFL teams, the chat will return incorrect information although ChatGpt has access to NFL.com lists. And of course, Apple was forced to neutralize its AI-based reconnaissance summaries after production Incorrect summaries of BBC News notifications.
Some AI proponents argued that sounding correct often means the text is A right but major problem with AI models seems to be that its users assume that everything they produce is right even when not. Google repeatedly presenting incorrect information during demos of its AI is emblematic of this problem. It seems that bosses at the technology giants are just freezing and shaking whatever their AI models produce without doing any work to bill it.
Meanwhile, Openai is dealing with other problems, namely a long-term legal battle with Elon Musk, who climbed this week after Openai’s early supporter made hostile 97 billion dollars to get the company AI Out of Altman. Although Musk has been able to rule Twitter despite all odds, this situation is considerably different, as Openai is private and remains non -profit. Although Openai is trying to convert into a non -profit company, as it now has no reliable duty to investors and probably does not have to consider acquisition bids.
