OpenAI, the company that developed the ChatGPT chatbot, introduced Operator — «agent» with artificial intelligence, which can be given a task, and it will independently perform it for you, using its own browser.
This is mentioned in OpenAI report.
As noted, «Operator» will be able to browse a web page and interact&with it, entering text, clicking and scrolling. It can be asked to perform a wide range of repetitive browser tasks, such as: filling out forms, ordering products and even creating memes.
It is currently only available in the US, for ChatGPT Pro subscribers, which costs $200 per month.
«This research preview allows us to learn from our users and the wider ecosystem, refining and improving the workflow. We plan to expand access to Plus, Team and Enterprise users and integrate these capabilities into ChatGPT in the future», — the company said.
«Operator» works on the basis of a new model called Computer-Using Agent (CUA). It is trained to interact with the graphical user interface — buttons, menus, and text fields that people see on the screen.
He can «see» (via screenshots) and «interact» (using all the actions that a mouse and keyboard can do) with a browser, allowing it to perform actions on the web without requiring special API integration.
If an AI encounters difficulties or makes mistakes, it can use its own self-correcting capabilities. When it gets stuck and needs help, it simply hands control back to the user.