As I notice more and more people use chatbots like ChatGPT and Claude as replacements for human companionship, I was struck by the hollowness that this new trend is. ChatGPT is not only not anyones friend (except perhaps Sam Altman's), it cannot be anyone's friend. There is nothing there to be friends with. No one is actually answering questions, no one is giving advice. There is no one there.

I am reminded of Blade Runner 2049, a world where preferring non-human companionship seems to have become just another sexual preference. Perhaps we are heading towards that world already. But we already know why a toy like ChatGPT seems attractive to those looking for companionship.

As a species, our ability to recognise patterns in nature has played a vital role in the creation of civilisation. The firsthumans saw patterns in weather change and in nature and carved lifestyes out of it that gave rise to migrations, agriculture, and eating habits. But they also saw patterns in the stars and in the behaviour of animals and these gave rise to religious superstitions.

I read recenty in Nicholas Carr' book The Shallows (affiliate link) that in the 1960s, Joseph Weizenbaum at MIT coded ELIZA into existence. It was software designed to talk to people like a chatbot and offer them counselling. Weizenbaum found that even small amounts of time spent with the software were enough for people to think of ELIZA as alive and intelligent and conscious. They reported that it could actually understand them. Even his own secretary, who had seen him write the code for ELIZA, fell into this mental trap.

ChatGPT is much more complex pattern than ELIZA and it is not surprising that people are failing to recognise that it is just code without any mind behind it. It's not becasue people are stupid. It's because people are people.