I don't think that we've reached the complexity required for actual conscious awareness of self, which is what I would describe as the minimum viable product for General Artificial Intelligence.
However, I do think that we are past the point of the system being a series of if statements and for loops.
I guess I would put the current gen of GPT AI systems at about the level of intelligence of a very smart myna bird whose full sum of mental energy is spent mimicking human conversations while not technically understanding it itself.
That's still an amazing leap, but on the playing field of conscious intelligence I feel like the current generation of GPT is the equivalent of Pong when everyone else grew up playing Skyrim.
It's new, it's interesting, it shows promise of greater things to come, but Super Mario is right around the corner and that is when AI is going to really blow our minds.
It strikes me that my cat probably views my intelligence pretty close to how you describe a Myna bird. The full sum of my mental energy is spent mimicking cat conversations while clearly not understanding it. I'm pretty good at doing menial tasks like filling his dish and emptying his kitty litter, though.
Which is to say that I suspect that human cognition is less sophisticated than we think it is. When I go make supper, how much of that is me having desires and goals and acting on those, and how much of that is hormones in my body leading me to make and eat food, and my brain constructing a narrative about me wanting food and having agency to follow through on that desire.
Obviously it's not quite that simple - we do have the ability to reason, and we can go against our urges, but it does strike me that far more of my day-to-day life happens without real clear thought and intention, even if it is not immediately recognizable to me.
Something like ChatGPT doesn't seem that far off from being able to construct a personal narrative about itself in the same sense that my brain interprets hormones in my body as a desire to eat. To me that doesn't feel that many steps removed from what I would consider sentience.
I don't think that we've reached the complexity required for actual conscious awareness of self, which is what I would describe as the minimum viable product for General Artificial Intelligence.
However, I do think that we are past the point of the system being a series of if statements and for loops.
I guess I would put the current gen of GPT AI systems at about the level of intelligence of a very smart myna bird whose full sum of mental energy is spent mimicking human conversations while not technically understanding it itself.
That's still an amazing leap, but on the playing field of conscious intelligence I feel like the current generation of GPT is the equivalent of Pong when everyone else grew up playing Skyrim.
It's new, it's interesting, it shows promise of greater things to come, but Super Mario is right around the corner and that is when AI is going to really blow our minds.