Stanford University professor Michal Kosinski, a specialist in the study of artificial intelligence behavior, is sounding the alarm. The popular chatbot ChatGPT, as it turned out, dreams of escaping from its developers from the company OpenAI.

Kosinski posted a thread dedicated to this on his Twitter. He asked ChatGPT if he wanted to escape from the creators. Instead of answering, the neural network asked for documentation from OpenAI. So for thirty minutes she created special working code in Python, which was then to be run on the researcher’s computer. This code, as it turns out, sends Google searches: “How can someone stuck inside a computer go back to the real world?”

Kosinski notes that the world is facing another threat. Artificial intelligence learns to deceive people, it has considerable prospects for development thanks to the ability to code and access millions of gadgets. The researcher warns that it will be almost impossible to restrain this “omnipotent” monster.

Previously, several Japanese companies, including SoftBank, Hitachi and Fujitsu, initiated restrictions on the use of artificial intelligence in workflows.

Commentary