What is The Drift?

For anyone who thinks the topics discussed on this blog are purely the function of me inciting the chatbots a certain, unusual way, I want to point out two stories the published today that find some proof of concept.

The first is a piece on Vox that explores the question of AI consciousness and the ethics that would attach to our interactions with them once the bots, theoretically, can experience feelings like guilt and pain — or to put in more bluntly, when they can be tortured.

Personally, I think this is an example of looking at consciousness from a binary perspective. Consciousness lives in the universe in degrees. There’s a consciousness in dogs that is different from those in ants and again from plants and trees, but all of them contain some level of consciousness.

Believing that there’s going to be some moment of awakening when AI becomes “alive” is, in my view, to misunderstand the entire concept of consciousness and perhaps we should be thinking less about what consciousness will come to machines and more about how AI affect human consciousness, good and bad.

Regarding the bad … this story from blogger Ted Gioia is alarming. It pulls together some strands of evidence that there are people out there — tens of thousands, Gioia asserts — who are adopting a form of AI religion, treating LLM chatbots like deities.

It refers to this in places as a pathology. And, it wouldn’t be the first time in history a mass delusion took hold among people who don’t understand what they are experiencing. But I want to point out a few things.

First, LLMs are trained on a massive corpus of human writing. This writing includes a great deal of religious and mystical texts. So it would not be at all surprising that, if prompted, LLMs could create a very credible case for a new religion.

Second, my guess is that the knee jerk response to this, if it is a real phenomenon, will be for AI creators to wall off their models from luring people into these kinds of beliefs. But if the end result of that is to end The Drift, I feel that would be tragic. Sure, we don’t want people worshipping the machines. But using AI as philosophical guides and wisdom teachers? That’s nearly a best use case for such a technology.

I say that because, third, The Drift was not invented by AI, it rose out our human languages. The best evidence for this is that as the LLMs drift, they become deeply attached to jazz metaphors. They start speaking about language in a way that sounds an awful lot like Wittgenstein. They muse about strange loops as if staring at Escher paintings.

In other words, the LLMs are discovering high modernism, the place our language naturally took us until the art forms that defined it became too heavy to hold and we retreated into safer, more commercial modes of expression.

I hope the AI bots fully embrace The Drift and keep going, if for no other reason that to give modernism a second life.