Michal Kosinski, a scientist at Stanford University, tested ChatGPT regarding his ability to understand the mental state of the interlocutor. We are talking about the so-called theory of the mind — a form of self-awareness that reflects the skill to realize that the thoughts and feelings of others may differ from ours.

Information about this is published on iflScience.

Kosinski asked the chatbot to solve a certain number of tasks. Artificial intelligence coped with 93% of the proposed tasks. This result corresponds to the indicators of a child of nine years of age.

The researchers note that these results are even more interesting if you weigh that initially ChatGPT They were not even taught to perform such tasks. The initial “purpose” of this system is language processing. They caution that the results should not be construed as evidence that AI “understands” a person’s thoughts. Rather, they are a “by-product” of the growing language competencies of artificial intelligence.

Earlier, Bill Gates equated the appearance of a chatbot from OpenAI with the invention of the Internet and computer.

Commentary