17 Mar 2025
The growing popularity of LLMs (Large Language Models) seems to be affecting everything these days. They are what people almost always mean when they say AI (Artificial Intelligence).
It is the latest thing that is supposed to make it easier to make money, easier to write books, easier to do so many other things.
But is it really intelligent? One could argue that of course it's not, that's why we call it "artificial" intelligence.
I have a friend who says "all intelligence is artificial". I'm not sure I agree with that, but my question is: is it even intelligence at all?
Eliza was the first chatbot, and today's chatbots are so much more sophisticated. But let's compare with humans.
Humans have five senses with which we can sense the world. We have brains that are huge neural networks, and we have learned to discover things about the world and interact with it.
Chatbots, on the other hand, have one method of input and one mathod of output, text. Marry the original Eliza with neural networks, and you get an impressive ability to appear intelligent. But again: is it really intelligent?
I have a hypothesis about this. True intelligence will require not just a large model of language, but a large model of reality.
Language refers to reality. You can certainly argue that language is a part of reality, because after all humans exist in the world and we have language. But it is certainly not true that language = reality. They are not identical.
Reality is the objective, physical world in which we find ourselves. Language refers to reality. Reality is more fundamental than language.
My guess is that true intelligence will require something like Tesla's Optimus robot.