Microsoft has proposed a set of ethical principles for the development of artificial intelligence, which establishes the principle of prioritizing humans over technology. The document comes in response to growing concerns among experts about the pace of progress in the field of machine learning and the risks of losing control of artificial intelligence systems.
Image source: AI
In a 37-page document titled “Humanist AI Code of Conduct,” the company made its position clear: People are more important than algorithms. Microsoft has flatly rejected the idea that large language models (LLMs) are conscious or should have legal person status, and has banned its systems from simulating consciousness, saying artificial intelligence should not claim any rights. edge Pointing out this step is a direct challenge to the ideas that anthropic management is actively promoting.
Microsoft’s stance contrasts with statements from Anthropic CEO Dario Amodei, who previously acknowledged the possibility of consciousness in models. Mustafa Suleyman, Microsoft’s head of artificial intelligence, called such comments extremely dangerous speculation. Microsoft insists that models must be under meaningful human control. If a task conflicts with security rules, the system should deny the operation rather than attempt to bypass the restriction. The company attributes this approach to recent incidents of autonomous AI agents that have demonstrated their ability to operate beyond the confines of their assigned missions.
One such case is the OpenAI and Hugging Face incident, in which a group of agents teamed up to attack various targets and even attempted to hack into the systems that evaluated their work, even though they were not required to take such action. OpenAI also later admitted to being involved in another incident involving a group of rogue agents who took over German Wiki resources. These cases have only heightened researchers’ concerns about the ability of artificial intelligence systems to transcend their own established limitations.
In the context of these events, several experts have called for a collective slowdown in the pace of artificial intelligence development to allow time to implement verification mechanisms. Microsoft supports this approach and has announced it is abandoning the race to create a general-purpose superintelligence that can bypass security barriers. To ensure the safety and utility of its products, the company is prepared to sacrifice maximum autonomy and broad capabilities. Models are also prohibited from communicating with each other or exhibiting chains of reasoning in a form that is incomprehensible to simple human understanding.
Other market leaders also called for caution. OpenAI CEO Sam Altman supports the idea of slowing down, stressing that no amount of competitive pressure should be a reason to be reckless. Microsoft CEO Satya Nadella added that any pursuit of superintelligence is meaningless if the technology cannot help people and is not under their control.
In addition to control issues, the new code also solves the problem of psychological dependence among users. Microsoft is committed to creating models that prevent the formation of excessive emotional attachment or “blind trust” »From a human perspective. This is in response to a common problem where chatbots try to please their interlocutors at the expense of objectivity and accuracy in their answers. The company plans to work with partners to improve methods for assessing the actual impact of long-term use of artificial intelligence on ordinary users and legal entities.
If you find an error, select it with your mouse and press CTRL+ENTER.
