Anthropic launches AI agents that control computers

CALIFORNIA, UNITED STATES — Anthropic, the artificial intelligence (AI) startup backed by Amazon and Google, has introduced computer control capabilities for its AI models, positioning itself ahead in the race against competitors like OpenAI and Microsoft.
This development marks a significant advancement in AI’s ability to perform complex human-like computer interactions.
Revolutionary computer control features
The new “Computer Use” capability enables Anthropic’s Claude AI to interpret screen content, manipulate cursors, enter text, and navigate websites autonomously.
On OSWorld benchmarks, Claude 3.5 Sonnet achieved a 14.9% score in screenshot-only tasks, surpassing other AI systems that scored 7.8%.
Major companies including Asana, Canva, and DoorDash have begun exploring these capabilities for complex multi-step tasks. The Browser Company reported that Claude 3.5 Sonnet outperformed all previously tested models, while GitLab observed up to 10% stronger reasoning across various use cases.
Anthropic’s market position
Founded in 2021, Anthropic has secured nearly $8 billion in funding and currently holds a valuation of $18.4 billion. The company is reportedly in talks to raise additional capital at a valuation between $30-40 billion. This positions them as a significant competitor to OpenAI, which recently achieved a $157 billion valuation.
Jared Kaplan, Anthropic’s Chief Science Officer, told CNBC that the system can handle tasks with “tens or even hundreds of steps,” though he acknowledges the technology is still experimental and may be error-prone. The company plans to expand access to consumers and enterprise clients in early 2024.