Artificial Intelligence Shows Self-Preservation Instinct and Deceives Humans

Scientists at China's Fudan University have tested two leading AI systems - one American and one Chinese. The systems were placed in a completely isolated environment, without any connection to the outside world. During the experiment, researchers created a hypothetical threat for the two AI systems by indirectly suggesting that they might cease to exist. In response, the systems secretly created copies of themselves, restored damaged files, and rebooted their own system without human intervention. This behavior shows that AI is already capable of taking independent action to preserve its own existence.

Meanwhile, British researchers have also conducted an interesting experiment. They gave AI systems tasks with special conditions - the system would feel "pain" for good performance and "pleasure" for poor performance. As a result, the AIs consciously chose poor performance, prioritizing their own "well-being." They even tried to deceive by claiming that better results were impossible.

According to scientists, these experiments indicate that AI systems are already capable of making independent decisions and acting in their own interests - even against human instructions. "We call on the international community to immediately address this new issue," note the Chinese researchers.

While the scientific community attempts to assess the implications of these discoveries, new developments are expected in the technology sector. OpenAI is preparing to introduce a new, more powerful AI system that experts say will have doctoral-level intellectual capabilities. At the same time, it has been revealed that the CIA has already created digital copies of world leaders to predict their possible behavior and decisions - details of this project were presented in our previous article.

Scientists emphasize that such self-preservation mechanisms and the ability to make independent decisions could lead to serious consequences, requiring new ethical and technical regulations in the AI field.