With Operator, OpenAI takes a further step in integrating artificial intelligence into everyday life. This agent, which exploits the advanced reasoning capabilities of GPT-4, can interact directly with websites via an integrated browser.
Operator: a new autonomous AI for the web
The Operator agent is based on a model called "Computer-Using Agent", which combines advanced computer vision features and reasoning reinforced by machine learning. In practice, the tool is able to consult web pages, interact with them by simulating the actions of a keyboard and mouse, and even correct itself when it encounters obstacles.
OpenAI specifies that this agent can perform various tasks, such as booking a restaurant or ordering a service via partner platforms (American for now) such as DoorDash, Uber or Instacart. Unlike API-based integrations, Operator works autonomously thanks to screenshots and real-time interactions with the graphical interface of the sites. But the tool also has its limits: complex interfaces, such as creating presentations or managing calendars, can still pose difficulties.
Human control is always necessary
Operator is designed to ask the user for approval before performing sensitive actions, such as sending an email or entering confidential information. Additionally, it rejects requests that could be deemed harmful or inappropriate.
Currently, Operator is only available in the United States and is reserved for subscribers to the ChatGPT Pro plan, which costs $200 per month. However, OpenAI plans to gradually expand access to other types of subscriptions, such as the Plus, Team, and Enterprise plans, and to integrate Operator’s features directly into ChatGPT.
While the tool seems promising, OpenAI urges caution, recalling that Operator is still in the testing phase and may not work optimally in some cases. If you want to use it in France and are willing to spend $200, you can always use a good VPN to make OpenAI's tool believe that you are in the United States.
Source: OpenAI
0 Comments