Anthropic has unveiled Claude Sonnet 3.5 — a new version of its AI model featuring computer control capabilities. Thanks to a new API called Computer Use, the AI can perform actions on a computer just like a human — moving the cursor, clicking buttons, and more. This feature is already available in a public beta, and companies such as Asana, Canva, and DoorDash are actively exploring its potential.
This new capability can expand the use cases for AI, enabling it to perform general-purpose tasks. Claude will be able to automate routine processes, conduct research, or even test software. “We taught Claude to understand what’s happening on the screen and use available tools to complete tasks,” the company’s official blog states. However, the developers still recommend starting to test the new functionality with less risky tasks while the service is under active development.
To prevent abuse, special classifiers have been developed to monitor risky actions and help avoid potentially dangerous situations, such as the spread of spam or misinformation.
Additionally, Anthropic has announced the upcoming release of an updated version of Claude 3.5 Haiku — the most cost-effective model in the Claude series, which will be available in the coming weeks. The new Claude 3.5 Haiku offers low latency, improved instruction following, and the ability to handle large volumes of data.
