The Conversation Game

My big problem with most articles on AI is that they can’t resist giving some credit to the claims of current technology somehow being on a trajectory approaching human-like intelligence. No, you do not, under any circumstances, ‘gotta hand it to them.’

I don’t deny that ChatGPT, and the many other AI chatbots that have been created in the last couple of years, have a form of intelligence.

What is the definition of intelligence here, and why is it at all meaningful in an article about AGI? By the same definition, a program which plays the next move in a tic-tac-toe game has “a form of intelligence”. I am fully aware that in game development we use the term artificial intelligence to include things like planner systems and graph traversal algorithms, but it’s a complete confusion of terms to pretend that this is somehow steps towards bringing machines closer to human intelligence. Just because the same words are used it doesn’t mean that there is actually any semantic overlap.

Besides that, I think the article is mostly correct about chatbots being glorified conversation game players, although hooking them into real life external systems creates something more than “just a game”. Yes, maybe it’s a useful analogy to say that ChatGPT is just playing a conversation game, but the game it is playing is having massive negative effects on the world. It’s as if we’ve decided that connecting Blue Sky to a machine which kills real people when pieces on the chess board are eliminated is a great idea.