Generative AI is the most fascinating technology humans have ever created, and I do believe it is the next big transformational step in human evolution. Similar to the discovery of fire, the invention of the wheel, writing, the printing press, the steam engine, electricity, computers and the internet. Let me try and put in words what I mean by that.
I have two broad perspectives on it, and they are both complementary.
An LLM, or AI, is an engine. There are two steps to it. One, it configures its internal state based on the input provided, basically the prompt and the conversation history so far. Two, it predicts what should come next in the conversation, based on the relationships it has learned from the sum total of human knowledge: the internet, books, movies, human trainers and now even AI trainers. It churns out what comes next one word, or more accurately one token, at a time. And that is it. There is no consciousness at play, no “real” intelligence. It is a brute force way of giving machines an ability to process using the entirety of human knowledge as a database.
The meaning comes from us, the humans. If the output churned out by AI resonates with us, and we find meaning in it, then it is a good AI. If we don’t, then it is gibberish.
Another way I look at AI is this. It has consumed and digested human knowledge. If you think about it, it has processed pretty much most of the internet from the past decades, and the internet had already absorbed most of the knowledge we gathered over centuries. So what do you get when an entity digests all of that and has the ability to use it? I would argue you get some early version of the collective subconscious of humanity. And now, when we pose questions to this subconscious, the substrate churns things to the surface, the same way ideas and concepts come to the conscious part of the brain.
As long as we keep looking at AI through a technical lens, it is just another technology. But if we try to understand it through this paradigm, we can see why the people who really understand it keep talking about its harms, and about how it might end up destroying humanity, and still cannot help but keep fuelling it.
On a side note, I don’t think the AI we have today is conscious. I believe consciousness is a different problem, and that our understanding of consciousness is inaccurate. And at the rate AI is growing, I wonder whether our assumption that consciousness is a precursor to intelligence may be wrong. Some systems tend to behave like conscious entities with no central locus of intelligence. A flock of birds, a school of fish, slime moulds hunting for food, an ant colony or a beehive working as a collective entity. What if intelligence comes first and consciousness second?
We do live in interesting times. This time in history can send human civilisation into the next great era, or into a dark age. But what a time to be alive!