OpenAI has introduced Operator, an AI agent capable of carrying out online tasks. Operator is available as a web application and can perform activities such as booking tickets or completing online grocery orders. It uses a model called Computer-Using Agent (CUA), built on OpenAI’s multimodal large language model, GPT-4o.

The tool is currently accessible at operator.chatgpt.com for ChatGPT Pro users in the United States, who pay $200 per month. OpenAI has stated that the service will become available to a wider audience in the future.

Operator has been developed to interact with websites using graphical user interfaces, such as buttons, menus, and text boxes, in a manner similar to how humans navigate online platforms. The model analyses the screen, executes an action, and reassesses the screen for further steps, breaking down tasks into smaller components and backtracking if necessary.

MIT Technology Review’s report, “OpenAI Launches Operator–An Agent That Can Use a Computer For You,” said the company tested and discovered that Operator outperforms similar tools from competitors. It surpasses Anthropic’s Computer Use and Google DeepMind’s Mariner in benchmarks designed to evaluate AI performance. On OSWorld, a test involving tasks like merging PDF files, Operator’s CUA scored 38.1%, compared to Computer Use’s 22.0%. Humans scored 72.4%. On WebVoyager, which measures browser task performance, CUA achieved 87%, ahead of Mariner’s 83.5% and Computer Use’s 56%.

Also Read :- Perplexity introduces AI assistant on Android, entering competition with Alexa and ChatGPT

Unlike its competitors, Operator runs on a remote browser hosted on OpenAI servers, allowing for simultaneous task handling. During a live demonstration, it was shown completing multiple tasks, such as booking a restaurant reservation via OpenTable, searching for event tickets on StubHub, and processing a handwritten shopping list for Instacart.

The system collaborates with businesses such as OpenTable, StubHub, DoorDash, Instacart, and Uber. It suggests specific websites for tasks but also allows users to provide direct instructions through a text box.

Safety tests have been conducted on CUA to ensure it avoids unacceptable actions. The model is designed to stop and request user input before proceeding with tasks that could lead to external consequences.

OpenAI intends to release CUA’s broader capabilities through an API, enabling developers to build their own applications. Operator is currently a research preview, which means its scope is limited to browser-based tasks and will evolve according to the feedback it receives from users.

Be a part of Elets Collaborative Initiatives. Join Us for Upcoming Events and explore business opportunities. Like us on Facebook , connect with us on LinkedIn and follow us on Twitter.

"Exciting news! Elets technomedia is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest insights!" Click here!