That's why this worries some

The interface between the chair and the computer is an artificial intelligence rather than a human… Interacting with the computer and its applications using a mouse and keyboard is what the new version by Claude Sonnet 3.5 from Anthropic can do. Evolution for the better and perhaps for the worse.

Driving our cars, answering mail, creating our photos and videos… coming soon, Amnesty InternationalAmnesty International He will be able to replace us in wide areas of our lives, for better and perhaps for worse. For the best, it is best to rely on the ethical and virtuous side of the human company developing Claude's chatbot. Its founders are former members of OpenAI who left due to the directions taken by the company. They are relying on the public interest of AI rather than immediate profit, with potentially disastrous consequences. Ironically, for worse, Anthropic has just published an improved version of Claudius 3.5 Sonnet. Her peculiarity is that she is able to understand and interact with anyone to requestto request office. In other words, AI can replace the “seat screen and keyboard” interface, i.e. the screen. Homo sapiensHomo sapiens.

Thus, thanks to the API called “ Use of computer “, the new model available in open beta can imitate keystrokes, button clicks and mouse gestures. To achieve this result, Anthropic trained its model to see and decipher what is happening on the screen via The catch. It then uses on-screen software tools to complete tasks. To move the mouse cursor to the correct place and click, Claude counts a number pixelpixel Vertical and horizontal. Ultimately, this gives the illusion of a real person sitting in front of him ComputerComputer.

See also  Jodorville. An internet point in the town hall

We can quickly imagine the usefulness of such a system in a “remote working” mode. Practical for pretending you are in front of the computer during a nap. We can also imagine a form of misunderstanding. With this ability, in front of the computer, Claude would likely be able to use ChatGPTChatGPTAnd Gemini and maybe himself…

Even more dangerous is that this independent interface could be a cause for concern if misused. Implemented by Hackers On a computer, it can enable vulnerabilities in applications to be directly exploited, and that's just one example. Without denying that the new Cloud Sonet 3.5 can be used for bad purposes, Anthropic highlights the advantages that can be obtained from its use.

AI keyboard and mouse

For them, it is the usage that will also determine what can be done with this tool. Using it will increase the security level accordingly. In all cases, humans remain in control by making specific claims. They are the ones who guide Claude's actions. This could be through examples such as ” It uses data from my computer and the Internet to fill out this form “, according to an interview with Anthropy's director conducted by the media TechCrunch. For example, Claude was tested for his ability to book or modify a flight. It was successful only in half of the cases. However, it is not uncommon for a person to also encounter difficulties with the online booking interface. But this is only the beginning..

If this is Claude's first time, the commands will be executed on a computer such as one used for browsing InternetInternet For example, it is not new inuniverseuniverse artificial intelligence. Some specific operations on the computer can actually be run through an interface based on GPT-4o, for example. But the anthropic model would be more robust and robust. He can also self-correct by trying to complete the task differently when he encounters obstacles. He also knows how to achieve goals that require hundreds of steps.

See also  Editorial positions are cut - Computer and Medien

However, for now, Anthropic admits that the new version of Claude has some difficulties with basic actions like scrolling and zooming and can also miss some actions and notifications due to its analysis system. ScreenshotsScreenshots. In short, if the principle exists, its smooth exploitation is elusive, and even less so when it comes to carrying out abusive acts. From a security standpoint, at the moment, restrictions prevent Claude from publishing on the site Social networksSocial networksCreate online accounts and interact with government websites. Although they want to minimize risks, these restrictions show a certain enthusiasm on Anthropic's part regarding the use that can be made through this interface.

Frank Mccarthy

<p class="sign">"Certified gamer. Problem solver. Internet enthusiast. Twitter scholar. Infuriatingly humble alcohol geek. Tv guru."</p>

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top