Anthropic is expanding its Claude models to include the ability to carry out tasks independently on the PC. The function is being tested by developers.

Tired of filling out digital forms? Wouldn’t it be easier if AI could do this for you? It’s not far away music. Anthropic announces new enhancements to Claude 3.5 Sonnet, one of the latest versions of the company’s LLM. Claude 3.5 Sonnet should now be able to work independently with a PC.

In a YouTube video, Anthropic shows what is possible. In the video we see how Claude processes a form completely independently. The AI ​​model itself searches for the requested data in various applications. It takes screenshots of your screen to find information and enter it into the appropriate field.

Cumbersome and error-prone

Anthropic is not yet putting this technology into the hands of a broad audience, but is letting a limited circle of developers test it. The company itself states that Claude’s computer skills are still very experimental and “sometimes cumbersome and error-prone”.

There are still many actions that you routinely perform on a PC that the Anthropic model is not yet capable of, such as: B. Drag and zoom. Anthropic has also deliberately introduced some restrictions. Claude is not given access to your social media accounts to prevent AI from writing messages on your behalf, and government websites are also marked off-limits.