Anthropic releases upgraded Claude 3.5 Sonnet AI model that can control your PC

Anthropic has introduced an upgraded version of its Claude 3.5 Sonnet model, capable of interfacing with desktop applications via a new "Computer Use" API. This feature, now in open beta, allows the AI to perform tasks by emulating keystrokes and mouse movements. Despite its advancements, the model struggles with basic actions like scrolling and has been shown to complete only half of certain tasks, such as flight bookings.

The model, described as an "action-execution layer," can browse the web and use any application, although developers are advised to explore low-risk tasks due to potential errors. While the AI demonstrates superior coding capabilities and self-corrects when encountering obstacles, its risk of misuse raises concerns. Anthropic emphasizes that monitoring its usage in real-world applications is crucial for understanding and mitigating risks associated with these technologies.

In addition, Anthropic plans to release Claude 3.5 Haiku, an affordable model with enhanced performance, suitable for generating personalized user experiences. Despite the competitive landscape of AI development, Anthropic continues to highlight the importance of safety and observation in deploying powerful AI models.

Source: TechCrunch