OpenAI is rolling out a new feature called Operator, allowing some ChatGPT users to utilize an AI-enabled web browser for various online tasks such as booking trips, shopping for groceries, and searching for deals. This AI tool harnesses a model trained on both text and images, enabling it to understand commands and navigate the internet seamlessly.
The introduction of Operator follows similar developments from competitors like Google and Anthropic, which have also launched AI agents capable of online activities. These AI agents represent a significant evolution in artificial intelligence, moving from simple inquiry-response systems to more proactive assistants capable of executing complex workflows. Peter Welinder, OpenAI’s VP of product, emphasizes that this transition is designed to enhance people’s productivity and improve the quality of their work.
However, the deployment of Operator comes with concerns. Streaming an AI-powered web browser may lead to unexpected behaviors. OpenAI acknowledges potential risks, including misinterpretation of commands and the chance for malicious manipulation. To mitigate these risks, the company has incorporated safety measures, ensuring that Operator checks with users before executing any irreversible actions.
OpenAI’s new tool will first be available as a research preview to ChatGPT Pro users, with the plan to expand access as the system is fine-tuned based on real-world usage. Demonstrations have illustrated Operator’s capabilities, including successfully booking an Amtrak trip and making restaurant reservations via OpenTable.
The AI operates using the GPT-4o model, equipped with enhanced training to execute tasks effectively within a browsing context. OpenAI is also working on making its Computer Use Agent accessible via API for broader use.
