Anthropic’s latest development allows its Claude AI assistant to control a computer like a human would. A feature called “Computer Use” entered public beta today. Using the computer, Claude can be directed to perform tasks such as “viewing the screen, moving the cursor, pressing keys and typing text,” according to the company. .
We’ve created an API that allows Cloud to understand and interact with computer interfaces.
This API allows Cloud to convert commands into computer commands. Developers can use it to automate repetitive tasks, conduct testing and QA, and conduct open research. pic.twitter.com/eK0UCGEozm
— Anthropic (@AnthropicAI) October 22, 2024
In theory, this could make AI even more useful in automating repetitive computer tasks. But just a second admitted that this application of Anthropic’s AI models, which focuses on computer usage, is still in development and, let’s say, buggy as heck. In an internal test, Claude stopped in the middle of an assigned coding task and started opening images of Yellowstone National Park, the company said. While this is unusual human behavior (who doesn’t want to take a break during the workday to look at natural beauty?), it’s a reminder that even the best AI models can make mistakes.
Anthropic has also released an improved version of it, in addition to demonstrating the use of the computer Model alongside a new model called the Claude 3.5 Haiku that will be launched in October. In August, Anthropic joined OpenAI to share its work with the US AI Security Institute.
