Operator: Meet the ‘Brother’ artificial intelligence of chatgtp for you

OpenAI, the owner of Chatgpt, launched a new artificial intelligence (AI) service this Thursday (23). This is the operator, a tool that can perform a series of tasks for you in a browser.

The operator is currently combined with the company’s most modern model and using a new computer (or CUA shortening), and has been created based on the GPT-4O language model that can even discuss with the user depending on the requests. .

This allows him to do: Interact with graphic interfaces such as buttons and menus on a website and perform basic or more complex tasks in a page. or application. In this way, the usability of the vehicle goes far beyond talking to you or searching on the Internet: it performs actions as if it were a personal assistant.

Other companies have already developed or initiated public tests with their own artificial intelligence representatives. This is the case of Anthropic, the owner of Claude Chatbot, and Google’s laboratory in the industry Deepmind. However, OpenAI guarantees that he performs experiments with competitors and that the operator performs better in all activities.

How does the operator work?

While the operator receives a command via text like other chat robots To understand the interface of the page in question, screen data pixel pixel works and activate the desired action. HE It uses a screenshot to have this context and understand the task as a multi -mod. – For example, he will only use something other than writing.

Then artificial intelligence It uses a virtual keyboard and cursor to navigate on websites, access links and even fill the forms. According to the data provided by the user. HE You can do all this without the need for API authorization Or something like that, like a person who reaches the addresses.

In one of the screening videos, the OpenAI employee asks the operator to search for a specific recipe on a special website and add certain materials to the shopping cart in the online store. All Browser actions are carried out in real time and the user can only watch While doing the whole job.

For example, the tasks used include: Spoon a table in a restaurant, buy tickets for a show, order for delivery and request a shipping vehicle per application.

Requests can be fully customized with user guidanceFor example, such as demands for materials that are not available at the desired meal or the best times for booking. This past is stored only for individual use and similar commands can be re -made with a single click in subsequent sessions.

Since this is a more advanced language model, He can take lessons from requests to improve his own performance and take into account previous actions.Especially when repeating tasks on the same website. The user should only intervene in certain situations, How to switch from Captcha authentication or how to enter the user name and password, This is sensitive data that is not caught by artificial intelligence.

Usability

now The use of operators is limited to the UK UNIVERSITY Users and specific site tasks. Initially, subscribers to the Chatgpt Pro plan, the most expensive version of the company’s paid plan, will be able to access the service.

Chatbot wandering for you It will be available in other countries and also for subscribers of other plans.Plus and corporate.

OpenAI also promised make us -using Operator CUA for more developers to create their own artificial intelligence intermediaries To perform more specific tasks in a browser. It will also continue to work on the vehicle and will enable the vehicle to perform more complex tasks in future updates.