Theory of Mind over Matter?

A set of tests is available to infer the capability of AI language models to discern human intents, beliefs and thoughts. Its called the Theory of Mind test and it has be run against all of the released versions of Chat GPT.

See this link for the paper explaining the test and its results which are based on a series of experiments first developed in 1978 to measure the complexity of a chimpanzee’s mind to predict the behavior of others.

Latest tests were run against ChatGPT (GPT 3.5) and they showed some startling increase in capability of the AI model. It was found that the latest model has the ability to understand text, puzzles and challenges like a 9 year old human. For example, changing the external wrapper on a bag of popcorn to show chocolate instead. The ChatGPT AI can figure out what the human response to this deceptive change will be.

This article also describes the tests in more detail and explains differences between earlier AI models and those produced in 2022.