To begin using it, however, you will be required to have a subscription to ChatGPT Pro.
OpenAI has started providing a sample of a new tool that allows users to navigate within a web browser. This tool is named Operator. Apparently, the program is driven by something that the corporation refers to as a Computer-Using Agent, as stated in a blog post that was published on Thursday. OpenAI describes the model as having been "trained to interact with graphical user interfaces (GUIs)," which are the buttons, menus, and text fields that people see on a screen. "CUA is trained to interact with GUIs in the same way that humans do," "Because of this, it is able to carry out digital tasks without relying on APIs that are specific to the operating system or the web."
On the basis of OpenAI's GPT-4o model, the most recent version of Operator was developed. The vision skills of the algorithm are combined with "advanced reasoning" that has been developed through reinforcement learning. The ability to "break tasks into multi-step plans and adaptively self-correct when challenges arise" is possessed by the operator. The next stage in the evolution of artificial intelligence, according to OpenAI, is represented by that capability.
Opening AI advises that Operator is "still early and has limitations," and that it will not "perform reliably in all scenarios just yet." This warning is similar to what OpenAI has said in previous research previews. For instance, the agent would gain tremendously from the user taking a few more moments to provide a more descriptive prompt. This is because the complexity of the activity and the interface involved all play a role in creating this benefit. According to The Verge, in the event that Operator ever becomes stuck on a task, it will hand control over to the user. Additionally, it will take control anytime a website requests critical information, such as login credentials, and it will do so automatically. The corporation claims that it developed the technology with the intention of "refusing harmful requests and blocking content that is not permitted."
Customers that subscribe to OpenAI's ChatGPT Pro service, which costs $200 per month, will be the first to have access to the Operator. Additionally, it is forming partnerships with businesses such as Instacart in order to provide the agent capabilities on their platforms; however, in order to test the integration, you will once again require a ChatGPT Pro subscription.
The number of artificial intelligence agents that are capable of navigating a web browser or a complete operating system is growing, and Operator is now one of them. The first company to provide this functionality was Anthropic, which did so in October with the launch of the Claude 3.5 Sonnet model. Google, on the other hand, did so more recently with the Gemini 2.0 model and Project Mariner.
