TLDR: I think the simple architecture brought unbelievable results. The way ChatGPT like AIs articulate is so aligned with reality and with what I observe that I have difficulty distinguishing the real from their simulation and I wonder if we can live in the same world as equal? Given that machines can perform very closely to humans I feel like the human brain is not so mysterious any more.
I recall a conversation with a great friend where he suggested that the AI hype is an overreaction by people who don’t really understand how rigid a system it is, and that its architecture will never make it a truly intelligent being. I instinctively knew that I disagreed with him, but at the time I did not know exactly how to express how he should perceive something like an Artificial Intelligence.
I have seen AIs evolve and make fewer mistakes over time, and I have always been impressed by their capabilities. I find the idea of giving all digital data produced by humans to a machine fascinating (mostly open data). Their rigid algorithms are able to mimic the way we write, even interchanging writing styles and structures. They pick up on things like prejudice and bias, they pick up the way we think and approach problems and situations.
ChatGPT passes modern variants of the Turing test and has fooled people into thinking they are talking with an actual human. Yet people don’t consider it a conscious being?
I’m not sure if I’m conscious myself. When I try to answer questions such as "Are you alive?", I say yes because I feel and I know so, yet if it is a machine saying that, for some reason we don’t think it’s true. I don’t believe I’m being conscious when I say "I know I am alive" because I can’t even express what "to know" or "alive" means. I use these terms out of reflex and not from a conscious thought. I know how to use those terms because I have seen other people use those expressions in many different contexts. The feeling that one is alive comes to us by reflex, not from actually understanding what it means. Like us, ChatGPT doesn’t truly know what words mean but knows where and when to use them.
When people describe hallucinations, limited memory from conversations, getting questions wrong, bias, or providing false or misleading information, I never know if they are describing humans or AIs. It's hard to say when things are equal to each other, usually in mathematics, if I have functions fand gI can say they are equal if f(x) = g(x). Making me believe that humans and AIs are quite similar because they produce very similar results. In parallel, both start with an empty brain and are fed with information over time. On one hand, parents try and give children good information and show them good behaviors example as education. On the other hand, scientists give highly filtered data for their models to learn and produce the best of the results.
On of the biggest differences is that Humans are incredibly efficient compared with machines. The human brain requires a very small amount of energy and space to run. To run an Ai we need huge data centers with thousands of CPUs that need a lot of electricity.
When people say that the most recent Ai's architecture is not good enough to build a true intelligent being, they don't appreciate that this "simple" method got so good results and is cognitively better than humans at many tasks. One of the most recent achievements is winning the first place 2025 International Mathematical Olympiad ahead of humans. We can probably say the same things about humans. They are super rudimentary because their brain is built by stacking little things that transport an electric signal in one direction only (neurons). I cannot wrap my head around how one can remember things and concept such as memory don't make any sense. It looks impossible that a brain composed of such "simple" architecture, functions so amazingly. The point I'm trying to make is if AI is not truly intelligent, neither are we.