Anthropic launches AI for automated coding clicks

Anthropic, backed by Alphabet and Amazon, unveiled upgraded AI models with a new feature for autonomously executing computer tasks, minimizing user input.

This innovative “computer use” function enables the AI to determine “where to move the mouse, where to click, and what to type” to accomplish complex tasks, as explained by Chief Science Officer Jared Kaplan. The company provides software developers with three versions of its Claude AI models, which are priced based on performance levels. The latest updates affect Sonnet, the mid-tier model, and Haiku, the entry-level option.

The newly released Haiku 3.5 can generate computer code that is “almost comparable” to the previously launched Sonnet version, according to Kaplan. CEO Dario Amodei has indicated that the most advanced model, Opus, will receive updates by the end of the year.

Currently, the computer use feature is exclusive to Claude 3.5 Sonnet and includes safeguards to prevent misuse for spam or fraud. Kaplan acknowledged that the AI is not infallible. Mike Krieger, co-founder of Instagram and now Chief Product Officer at Anthropic, expressed the company’s desire for feedback from business customers to refine this feature. A dedicated lab team is also investigating how to extend this capability to consumers, which Krieger personally supports.

This feature is designed for software developers, signifying a shift towards AI agents—programs that can perform multi-step actions with minimal human intervention. Researchers view these agents as the next step in AI evolution, surpassing traditional chatbots that only generate text or code. Anthropic showcased the feature by coding a basic website and planning a sunrise outing using tools like Google Search and Apple Maps.