نموذج Anthropi الذكاء الاصطناعي يمكنه التحكم في الكمبيوتر بدلًا منك

Anthropic's AI Model Can Control Your Computer Instead of You

A new generation of AI agents is emerging, capable of taking control of your computer and performing tasks autonomously.

Anthropic, a company backed by Amazon and a competitor to OpenAI, has released a new version of its Claude 3.5 Sonnet model. This model has the capability to perform a range of basic desktop tasks, such as entering keystrokes and mouse clicks, allowing it to utilize any applications you have installed.

The Rise of AI Agents: Beyond Chatboxes

This update signifies Anthropic's entry into the industry race to push commercial AI models beyond chatbot limitations and transform them into full-fledged “AI agents.”

What are AI Agents?

The term “AI agents” refers to AI models designed for productivity, aiming to utilize software and perform computer tasks like humans, with varying degrees of complexity. While some agents, such as Devin from Cognition AI, are specifically designed for programming, Anthropic promotes its new agent as versatile, claiming it can browse the web, use any website or application, and perform a wide range of tasks, from technical programming to simpler activities like trip planning.

Demonstrating Capabilities

In a demonstration, Claude was instructed to plan a trip to see the Golden Gate Bridge at sunrise with a friend. The AI opened a web browser, searched for a suitable viewing location on Google, gathered additional details, and added the trip to a calendar app. In another demonstration, Claude was tasked with creating a simple website, which it accomplished using Microsoft's Visual Studio Code. It even opened a local server to test the website it created. Despite a minor error in the creation process, the AI corrected the code when requested.

Reliability Concerns and Potential Risks

While promising, AI models still struggle with reliability, particularly when it comes to coding. Even in a simple test involving booking flights and modifying reservations, Claude 3.5 Sonnet successfully completed less than half of these tasks.

This raises concerns about letting this experimental and sometimes unpredictable technology roam through your computer files and use your web browser. The potential risks and ethical implications of granting such extensive control to AI agents require careful consideration and thorough testing before widespread adoption.