OpenAI has unveiled Operator, an AI assistant that can perform various tasks for users as specified through a web browser.
Operating through the user’s own web browser, Operator has the ability to comprehend content displayed on websites and interact accordingly, such as typing, clicking, and scrolling – a method that sets it apart from bot tools. Moreover, it can fill out forms, place orders, and more.
Operating on OpenAI’s new model called CUA (Computer-Using Agent), Operator combines the visual capabilities of GPT-4o with reasoning through reinforcement learning, enabling it to respond to on-screen elements just like a human does and interact with them.
When users command Operator to perform tasks on a website, they can take control themselves as needed, especially for entering sensitive data like logins, payment information, or CAPTCHAs. Users can also customize profiles for each website if there are varying sets of information required.
Operator is now available for Pro customers in the United States, with its functionality still under research preview, which may include potential errors and limitations when working with certain websites. Expansion to other user groups is planned for the future.
TLDR: OpenAI introduces Operator, an AI assistant that performs tasks through a user’s web browser, utilizing a combination of visual understanding and reinforcement learning to interact with website content like a human. Its availability is currently limited to US Pro customers, with potential for wider access in the future.
Leave a Comment