Comment by low_tech_love

This is a very interesting conversation. Correlates well with the responses I got from the colleague during the meeting. Would you ask ChatGPT to do a t-test for you and blindly accept its results as well, regardless of whether the math behind it was sound or not? The reason why we use math and statistics in experimental research is because we want objective results, not simply results that correlate with our expectations (that we can get from watching YouTube or reading blogs). The objectivity of K-Means buys me the trust that whatever clusters I get have been obtained with a well-know and understood method, in which my expectations have absolute no influence. Also, I know that the next person will get similar results, which also gives me trust in their results. So we can all have a shared, independent, objective understanding of a piece of data.

I wonder, if well-educated and technically-literate people like him and you are willing to accept arbitrary results from a language model as a replacement for objective math, then what should we expect from the general public?