The success of ChatGPT has made OpenAI the hottest star start-up right now. However, competition in the AI space is heating up, with a variety of chatbots emerging. In the face of competitors such as Google and Claude, OpenAI is also stepping up the development of other products.
According to **, OpenAI is developing two revolutionary agent software. One of them is capable of effectively taking over the customer's device to automate complex tasks. For example,Transfer data from documents to electron** for analysis, or automatically fill out expense reports and enter them into accounting software, or create an itinerary or book a flight ticket based on a certain budget.
According to a person familiar with the matter,**Ability to automate clicks, cursor movements, text input, and more, just like a human operating software.
The CEO of OpenAI has frequently mentioned his vision that ChatGPT will become a "super-intelligent personal assistant" for work, paving the way for the all-powerful AGI (general artificial intelligence that can do everything a human can do).
However, OpenAI also has to contend with users' privacy concerns about computer takeovers and automated operations. Takeovers may conjure some people up to malware, which is known to illegally take control of computers and steal data.
According to the report, according to OpenAI's plan, the agent that takes over the user's computer needs to be licensed to work, and in order to ensure the response speed, OpenAI may need to partially run a large model locally (the so-called "edge computing"). In contrast, to access ChatGPT now, all computing is done in the cloud via Microsoft Azure.
*OpenAI's agent product has been in development for more than a year, and it is unclear when it will be released.
Last month, an OpenAI employee posted a job post on X and said the company's upcoming new product could "change everything."
The company's announcement also revealed clues that OpenAI wants to launch an agent product. At a customer event in November, OpenAI unveiled the Assistants API, which allows developers to build agent-like experiences in their own applications, borrowing from APIs that allow GPT large models to generate graphs, track conversation history, and retrieve information from external documents.
However, while this tool allows developers to connect their agents to the Internet or perform tasks in designated external applications, it does not give the agent full control or knowledge of the user's computer.