top of page

Anthropic's Claude 3.5 AI Introduces Game-Changing 'Computer Use' Feature

23/10/24

By:

Piyush Sharma

New update allows AI to control a computer like a human, offering expanded potential for developers.

New update allows AI to control a computer like a human, offering expanded potential for developers.

Anthropic has launched a groundbreaking new feature in its Claude 3.5 Sonnet AI model: the ability to control a computer autonomously. Named "computer use", this feature enables Claude to observe a computer screen, move the cursor, click buttons, and type text, similar to human interaction. Currently available in public beta through the API, this update aims to revolutionize how AI can assist with tasks by directly interacting with devices, though it's still in an experimental phase.

This feature puts Anthropic’s Claude alongside AI tools like Microsoft’s Copilot Vision and OpenAI’s desktop apps, which also have screen observation capabilities. However, Anthropic's release goes a step further by enabling Claude to actually perform actions on the screen rather than simply observe.

While the computer use feature is exciting, Anthropic warns it can be “cumbersome and error-prone” at times. The AI pieces together screen snapshots to perform its actions, meaning fast-changing notifications or brief pop-ups might be missed. In addition, Claude is designed to avoid engaging in tasks like social media interaction, registering domains, or engaging with government websites to prevent misuse.


Aside from the headline feature, the Claude 3.5 Sonnet update boasts improved performance in coding and tool use. It scores higher on benchmarks like SWE-bench for coding and TAU-bench for agentic tool use, surpassing other publicly available models, especially in retail and airline domains.

Anthropic's computer use feature represents a leap toward AI-powered task automation, enabling developers to experiment with new levels of AI interaction. With rapid improvements expected, it could soon reshape how we use AI for computer-based tasks.

Key Takeaways:

  • Anthropic’s Claude 3.5 can now interact with a computer, using screen, cursor, and keyboard controls.

  • Currently in public beta for developers, the feature allows AI to perform tasks on a Mac or PC.

  • Although experimental, this update positions Anthropic's AI among other leading tools like Copilot Vision and ChatGPT desktop.

  • Claude’s coding and tool use abilities have significantly improved, outperforming previous models.

This could be a transformative step in how AI assists with everyday computing.



All images used in the articles published by Kushal Bharat Tech News are the property of Verge. We use these images under proper authorization and with full respect to the original copyright holders. Unauthorized use or reproduction of these images is strictly prohibited. For any inquiries or permissions related to the images, please contact Verge directly.

Latest News

13/12/24

Apple’s New HomePod Mini and Apple TV Expected in 2025

Enhanced with Apple’s proprietary “Proxima” chip for improved connectivity and smart home integration

13/12/24

Google’s Vision for Android XR: Bringing Smart Glasses and Headsets to Life

The Android XR platform aims to redefine augmented and mixed reality, powered by Gemini AI and seamless integration.

13/12/24

Google Launches Gemini 2.0: Ushering in the AI Agentic Era

The advanced multimodal AI model can generate images, audio, and promises groundbreaking agent capabilities.

bottom of page