Meet Operator: The New AI App from OpenAI That Can Help You Online
OpenAI recently previewed what its upcoming agent, called Operator, will be able to navigate the web and perform tasks on your behalf. The agent uses your browser to access and interact with any web page and will launch first in the US for users who subscribe to ChatGPT Pro.
Operator is based on the “computer-friendly agent” model that combines the vision capabilities of the 4.0 model with “advanced reasoning through reinforcement learning” to interact with graphical user interfaces (GUIs). Thanks to this integration, the agent becomes able to understand and process visual elements through screenshots, make decisions independently, and use actions allowed by the mouse and keyboard with the browser.
The operator will have its own awareness so that it can self-correct and give control to the user if it crashes. Additionally, there are other situations where it might also give up control, such as when a website asks you to enter personal data or login credentials for a website. To make matters worse, the operator will not always work autonomously because, for example, to send emails, it will have to ask for permission, among other situations.
Operator is expected to be part of the Plus, Team, and Enterprise plans as well, as it is currently only available to ChatGPT Pro subscribers. Additionally, OpenAI wants to integrate Operator into ChatGPT in the future.