A thinking machine

I’m going to preface all this with saying I’m not a biologist, data scientist or AI researcher. This is just, as the rest of the blog, my thinking process.

A friend and I had a discussion about a year ago what constitutes intelligence. It was in the wake of GPT 3.5, which spurred a lot of misapplied and half baked solutions of LLMs.

We came to the point where the question was if GPT takes input and produces an output (a response), isn’t that also what human beings are doing? We react to input from the world. And if GPT does this in a convincing way, isn’t that intelligence? And maybe we are exaggerating the human intelligence?

My friend leaned toward ”yes”, I was very much “no”. But I couldn’t motivate it other than it didn’t feel intelligent.

The other night, when I was sleepless in bed, it hit me. GPT only produces output when you feed it an input. I was thinking without input.

And so now I’m very curious if there is a machine that runs without input. If you don’t feed it, will its CPU¹ still run?

I’m relating this to humans, dogs and insects. I think of humans as intelligent; we think even if we are idle. My perception is that dogs do that as well. They continuously process. Insects are more input driven. It’s cold – they move. They are low on energy – they eat. (Simplified of course) This is where I put current LLMs, at least those that I have tried. That doesn’t mean that insects are not totally fascinating, or that LLMs are useless.

That will be my new definition of a thinking machine; processing without input.