1. News
  2. AI
  3. OpenAI Launches Operator: Your New AI Sidekick!

OpenAI Launches Operator: Your New AI Sidekick!

featured
Share

Share This Post

or copy the link

On Thursday, OpenAI unveiled its inaugural artificial intelligence (AI) agent, named Operator. Initially accessible as a research preview, this agent features a dedicated web browser designed for a variety of online tasks. It operates autonomously based on user prompts and is capable of booking tickets, making restaurant reservations, and purchasing products online. Currently, Operator is exclusively available to ChatGPT Pro subscribers in the United States, with plans for broader access in future subscription tiers.

OpenAI Launches Operator AI Agent

In a live stream, OpenAI CEO Sam Altman introduced the new AI agent, explaining its significance. Altman described AI agents as systems that can operate independently once assigned tasks by users, stating, “We think it will be a big trend in AI.”

operator ai agent OpenAI Operator

The Operator AI agent interface
Photo Credit: OpenAI

 

Operator is built on the Computer-Using Agent (CUA) model, which merges the visual capabilities of GPT-4o with advanced reasoning, as detailed by OpenAI in a blog post. The agent was refined through reinforcement learning and is able to interact with graphical user interfaces, such as buttons and menus. Its dedicated browser allows it to perform tasks while maintaining user focus on the screen.

The agent processes both text and images as input, utilizing raw pixel data to enact tasks through virtual keyboard and mouse actions. OpenAI emphasizes that Operator can manage multi-step tasks, adapt to unexpected challenges, and handle errors effectively.

Use Cases of the Operator AI Agent

Rowan Cheung, founder of the AI newsletter The Rundown AI, gained early access to Operator and shared various use cases in a series of posts on X (formerly Twitter). He noted that the AI agent effectively planned a weekend trip by gathering insights from Reddit, based on a specified budget and personal interests. Notably, when it encountered restrictions accessing Reddit, the agent successfully switched to a Bing search using ‘Reddit’ as a keyword.

In a different scenario, Cheung tasked Operator with researching noteworthy cryptocurrency tokens. When the AI encountered a CAPTCHA asking, “Are you human?” it promptly notified the user to verify their identity. Once confirmed, the AI re-assumed control to complete the task.

The design of Operator allows the user to intervene at any moment to adjust or edit the ongoing task and subsequently return control back to the agent when necessary, providing users with constant oversight.

OpenAI has also revealed collaborations with companies like DoorDash, eBay, Instacart, and Uber to ensure that Operator adheres to the respective terms of service agreements while accessing these platforms.

Operator’s Safety Risks and Mitigation

On the topic of safety, OpenAI has reported extensive testing to address three categories of risks: misuse, model errors, and frontier risks.

To combat misuse, the CUA model has been programmed to reject harmful requests and any illegal activities. The company has placed restrictions on gambling, adult content, and websites selling drugs or firearms. OpenAI has instituted both automated and human reviews of user interactions to further enhance safety.

Regarding model errors or inaccuracies, the AI agent is designed to seek user confirmation before finalizing any actions with potential consequences. Operator is programmed to refrain from assisting with banking activities, and it requires user oversight when accessing sensitive websites.

Frontier risks pertain to unforeseen behaviors from advanced AI models, which are often not subjected to comprehensive testing. OpenAI asserts that the CUA model has been assessed in accordance with its Preparedness Framework, with the Operator System Card detailing the safety measures in place and ongoing enhancements.

Operator is currently accessible at the operator.chatgpt.com URL for ChatGPT Pro subscribers in the US. OpenAI plans to roll out integration of the AI agent across all ChatGPT clients in the future, with a subscription to ChatGPT Pro priced at $200 per month (approximately Rs. 17,200).

OpenAI Launches Operator: Your New AI Sidekick!
Comment

Tamamen Ücretsiz Olarak Bültenimize Abone Olabilirsin

Yeni haberlerden haberdar olmak için fırsatı kaçırma ve ücretsiz e-posta aboneliğini hemen başlat.

Your email address will not be published. Required fields are marked *

Login

To enjoy Technology Newso privileges, log in or create an account now, and it's completely free!