I think the issue is that when people hear “AI,” their minds immediately jump to the sci-fi AI systems depicted as as smart or smarter than humans. They then see the stupid mistakes LLMs make and reasonably conclude these systems are nothing alike, so LLMs don’t count as AI in their minds.
However, the AI systems in sci-fi aren’t just intelligent - they’re generally intelligent. That’s what LLMs lack.
The way I see it, there are levels to intelligence. A chess bot is a narrowly intelligent system. It’s great at one thing but can’t do anything else. Then there’s Artificial General Intelligence (AGI), which is basically human-level intelligence. The next step up is Artificial Superintelligence (ASI) - a generally intelligent system that’s superhuman across the entire field of intelligence, unlike a chess bot that’s only “superhuman” at chess.
I’d say LLMs are somewhere between narrow intelligence and AGI. They can clearly do more than just generate language, but not to the extent humans can, so I wouldn’t call them generally intelligent. At least not yet.
And yeah, I don’t think sentience necessarily needs to come along for the ride. It might, but it’s not obvious to me that one couldn’t exist without the other. It’s conceivable to imagine a system that’s superintelligent but it doesn’t feel like anything to be that system.
Thank you.
I think the issue is that when people hear “AI,” their minds immediately jump to the sci-fi AI systems depicted as as smart or smarter than humans. They then see the stupid mistakes LLMs make and reasonably conclude these systems are nothing alike, so LLMs don’t count as AI in their minds.
However, the AI systems in sci-fi aren’t just intelligent - they’re generally intelligent. That’s what LLMs lack.
The way I see it, there are levels to intelligence. A chess bot is a narrowly intelligent system. It’s great at one thing but can’t do anything else. Then there’s Artificial General Intelligence (AGI), which is basically human-level intelligence. The next step up is Artificial Superintelligence (ASI) - a generally intelligent system that’s superhuman across the entire field of intelligence, unlike a chess bot that’s only “superhuman” at chess.
I’d say LLMs are somewhere between narrow intelligence and AGI. They can clearly do more than just generate language, but not to the extent humans can, so I wouldn’t call them generally intelligent. At least not yet.
And yeah, I don’t think sentience necessarily needs to come along for the ride. It might, but it’s not obvious to me that one couldn’t exist without the other. It’s conceivable to imagine a system that’s superintelligent but it doesn’t feel like anything to be that system.