Well put, and I agree with your general approach/skepticism! That said;
1. I don’t think we need to meet “always can improve itself”; rather, the contention is that there’s a high chance that there’s lots of improvement left to be had. An empirical claim rather than a theoretical one, in other words. The simple fact that humans are evolved creatures backs this up in spades, IMO — we’re still improving our own cognition by leaps and bounds using institutions, tools, and methods, and I don’t see any reason why that same dynamic wouldn’t apply to artificial cognitive systems.
2. I think dodging “intelligence” is exactly what he’s trying to do by listing concrete behavioral/cognitive differences. “Intelligence” is pretty much a useless term in science IMO, as was best expounded by Turing in his seminal 1950 paper, Computing Machinery and Intelligence:
People remember that paper as “you can tell a real AI when it can trick you”, but that’s not what he was trying to say at all; rather, he was trying to highlight that there is no such thing as a “real” AI, or “real” thinking, or “real” intelligence — just behavioral similarities and dissimilarities.
If the second bit grabs you/anyone, definitely watch some Chomsky lectures on YouTube about cognition. He centers his analysis on this, pejoratively calling discussions about “person”, “intelligence”, “thinking”, etc. mere terminological disputes, not specific enough to have much scientific value. His old refrain is a great one: does an airplane fly? Does a submarine swim? Kinda, if you want!
1. I don’t think we need to meet “always can improve itself”; rather, the contention is that there’s a high chance that there’s lots of improvement left to be had. An empirical claim rather than a theoretical one, in other words. The simple fact that humans are evolved creatures backs this up in spades, IMO — we’re still improving our own cognition by leaps and bounds using institutions, tools, and methods, and I don’t see any reason why that same dynamic wouldn’t apply to artificial cognitive systems.
2. I think dodging “intelligence” is exactly what he’s trying to do by listing concrete behavioral/cognitive differences. “Intelligence” is pretty much a useless term in science IMO, as was best expounded by Turing in his seminal 1950 paper, Computing Machinery and Intelligence:
https://courses.cs.umbc.edu/471/papers/turing.pdf
People remember that paper as “you can tell a real AI when it can trick you”, but that’s not what he was trying to say at all; rather, he was trying to highlight that there is no such thing as a “real” AI, or “real” thinking, or “real” intelligence — just behavioral similarities and dissimilarities.
If the second bit grabs you/anyone, definitely watch some Chomsky lectures on YouTube about cognition. He centers his analysis on this, pejoratively calling discussions about “person”, “intelligence”, “thinking”, etc. mere terminological disputes, not specific enough to have much scientific value. His old refrain is a great one: does an airplane fly? Does a submarine swim? Kinda, if you want!