Anthropic takes a new step in the development of artificial intelligence. His assistant, Claude, can now use the computer as a human would. An innovation that raises many hopes and questions.
Computer Controlled Artificial Intelligence: Claude's Artificial Intelligence, developed by AnthropicNow, he can control the computer independently, with a simple text request. You move the mouse cursor, click buttons, and type text as a human would.
This technology, called “computerization,” allows artificial intelligence to perform desired actions, such as filling out forms automatically by searching for the necessary information in Excel files or on the Internet.
Capacities are still limited
the first Tests However, it reveals some limitations. AI still has difficulty performing basic actions like page scrolling or zooming. The service is also paid, with billing for each task.
What's even more worrying is that Documents The technology suggests that artificial intelligence can sometimes follow instructions in images or on websites, even if they conflict with user requests.
Security risks
This innovation raises concerns in the field of cybersecurity. Follower Experts Fear that this technology could be misused to automate computer attacks or spread viruses on a large scale.
To reduce this RisksTesting must be performed in a virtual machine, which is an isolated virtual environment without access to sensitive computer data.
Paradigm shift
The real break is not technical, but about our relationship with trust. Unlike traditional tools whose operation is predictable, generative AI produces variable and uninterpretable results.
This paradoxical situation, where creators themselves call for caution while encouraging adoption of their technology, raises questions about the trust users can place in these new tools.
Pascal and Asmar
