OpenAI Competitor, Anthropic, Says Its New AI Model Can Control Your Computer
Anthropic, a firm financed by Alphabet and Amazon.com, introduced two revised artificial intelligence models on Tuesday, as well as a new capability to do computer tasks autonomously and preserve users’ keystrokes.
On Tuesday, Anthropic published an updated version of its Claude 3.5 Sonnet model, which can understand and interact with any desktop software.
The model can simulate keystrokes, button clicks, and mouse gestures using a new “Computer Use” API that is already in open beta, effectively simulating a person sitting at a computer.
“I think we’re going to enter into a new era where a model can use all of the tools that you use as a person to get tasks done,” says Jared Kaplan, chief science officer at Anthropic and an associate professor at Johns Hopkins University.
Kaplan presented WIRED a prerecorded demo in which a “agentic”—or tool-using—version of Claude was asked to assist in planning a trip to view the sunrise at the Golden Gate Bridge with a buddy. In response to the instruction, Claude opened the Chrome web browser, searched for pertinent information on Google, including the best viewing location and time to arrive, and then used a calendar app to make an event to share with a friend.
Anthropic provides software developers with three versions of Claude, a family of AI models, at varying price points based on performance.
This week’s updates are for Sonnet, the mid-tier model, and Haiku, the lowest.
According to Kaplan, the new 3.5 Haiku can generate computer code that is “almost comparable” to the Sonnet version published in June.
We earlier reported that Anthropic has announced the release of prompt caching.
This new feature improves the response times of AI’s large language models by allowing developers to pass them longer, more detailed prompts.
