The latest Anthropic AI agent can control your computer on your behalf. It will emulate a person sitting at a PC by imitating keystrokes, clicking buttons, and performing other functions.
The OpenAI competitor Anthropic named the feature “Computer Use,” which runs via the Claude 3.5 Sonnet AI model.
Developers may try the public beta via Anthropic’s API, Amazon Bedrock, and Google Cloud’s Vertex AI platform.
What the Anthropic AI agent means for the future
Having an artificial intelligence (AI) that can control computers has huge ramifications for global tech adoption. Arguably, the biggest potential impact is having AI agents that could work as your secretary or digital twin.
For example, you could ask your AI agent to book your flights and appointments. In response, it will navigate through your online accounts and websites to schedule your workday.
“We trained Claude to see what’s happening on a screen and then use the software tools available to carry out tasks,” Anthropic wrote to TechCrunch.
READ: How to use Anthropic’s Claude 2 chatbot
“When a developer tasks Claude with using a piece of computer and gives it the necessary access, Claude looks at screenshots of what’s visible to the user…”
“… then counts how many pixels vertically or horizontally it needs to move a cursor in order to click in the correct place.”
Nevertheless, humans remain in control by choosing specific prompts that direct Claude’s actions.
For example, a command may say, “Use data from my computer and online to fill out this form.”
“People enable access and limit access as needed. Claude breaks down the user’s prompts into computer commands to accomplish that specific task.”
READ: Anthropic releases Claude AI on Apple App Store
However, the AI firm admits that the Anthropic AI agent can be “cumbersome and error-prone.”
Consequently, “we’re releasing computer use early for feedback from developers and expect the capability to improve rapidly over time.”
The Anthropic AI agent also cannot access social media, preventing it from engaging in “election-related activity” and entering government websites.
Nevertheless, tech research firm Gartner says AI agents could make it easier for companies to earn returns from their AI investments.
The latest Capgemini Research Institute survey says 10 percent of organizations already use AI agents. Moreover, 82 percent will integrate them within the next three years.