Openai He showed a preview of what his agent of ChatGPT called Operatorthat is capable of navigate y Make indicated by users. This new function uses its own browser to access any page and interact with it, and will be released first in USA For those users signed to Chatgpt forthe new premium plan of 200 dollars a month.
Operator is based on a model of “Computer Use Agent” that combines the vision capabilities of model 4.0 with a “Advanced reasoning through reinforcement learning” in order to interact with the GUI (User Graphic Interfaces). Thanks to this integration, the agent is Able to understand and manipulate visual elements through screenshots, Making decisions autonomously and using the actions that allow a mouse and a keyboard with a browser.
Operator will have his own awareness to be able to self-bring and leave the user control In case of getting stuck. In addition, in other situations it will also give control, such as When a website requests that the user enter personal data or the login credentials on some website.
It will not always work autonomously, since, for example, To send emails should ask for permissionamong other situations.
Openai also announced that he is working with companies such as Dordash, Instacart, Openable, Priceline, Stubhub, Thumbtack and Uber for Operator to cover real world needs. However, Openai also commented that his agent has somewhat more complex interface problemssuch as the creation of presentations or calendar apps.
The future of Operator is also part of the Plus, Team and Enterprise plans, since at the moment it is only available for the subscribers of Chatgpt Pro. In addition, Openai wants Operator to be integrated into chatgpt in the future. While reaching the other plans and more countries, Openai continues to improve its chatbot with the introduction of the Projectsa new way of organizing chats.
What are artificial intelligence agents
Technological giants Openai, Microsoft, Google y Salesforce They are now accelerating the development of a new generation of artificial intelligence based on Agents.
These systems will mark a before and then in fields such as health, robotics and video games, Leaving behind the era of chatbots To give way to tools capable of making complex tasks autonomously.
These programs are developed to perceive your environment and make automatic decisions using artificial intelligence models. Therefore, it is not an AI with which a user can interact as chatgpt, but programs designed to perform tasks based on their environment.
And You have an agent You can use different methods to link with your tasks, always depending on the objective with which it was designed. Sometimes, you can do it with people by written text, or a series of questions to have a greater understanding of the context.
But it is also possible that it is on a technological device equipped with various types sensors that allow them to analyze their surroundings, similar to how an intelligent thermostat adjusts the temperature or a vacuum cleaner Roomba Learn the disposition of a room.
Experts distinguish three levels of sophistication in these agents. The evolution is fascinating and progressive: from the Simple reflex agents as Thermostats through the based on objectives as a romba, until you reach the most advanced based on utility capable of Weigh risks and benefits Before making decisions, even considering objectives that can go into conflict.