This paper is historically very interesting. However, I think (and I'm not alone) that it set AI research off in a bad direction. It's not that formal logic m

This paper is historically very interesting. However, I think (and I'm not alone) that it set AI research off in a bad direction. It's not that formal logic manipulation toward common sense is not important, but rather that it is not the skeleton key to the artificial intelligence dream. It's hard for me to _prove_ that statement, but we can see the palpable disillusionment of researchers who've tread this path. If I were personally doing AI research now, I would begin with emotions. Humans act because they want something, and they only use logic as a means to an end -- and certainly not perfectly at that. If our goal is to achieve a thinking program, a rational approach would be to walk down the course evolution has found -- simulate the mind of a fly, a toad, a dog, and a monkey first. It may be tempting to design something top-down, but that presumes we understand the top very well, and I don't think we do. (By "top-down" I mean from behavior to code; I'm sidestepping the biochemistry of neural activity, whose study implies we need to know the brain's hardware before we can emulate its software.) As a side note, some folks use AI and ML (machine learning) interchangeably, but there are really two goals around. I see McCarthy as wanting to create a program that could pass the Turing test (AI), while support vector machines and other ML tools are designed to perform advanced statistical analysis of data, along the lines of extrapolating a pattern. Related but different goals, and this paper has been viewed historically as about AI.