Yea, I think we might agree there but I don’t think that supports the original assertion that human thought is nothing but an (exceedingly complex) algorithm. You can also represent human thought as a system of hydraulic pressures, that’s what early psychology did, and how we got words like repression. But just because you can do that, and maybe even gain some useful knowledge from it - doesn’t mean actual human thought is actually made up of a complex system of pressures/valves - or algorithms. Your map may seem useful, but it ain’t the territory, is what I’m trying to get at, I guess.
To be clear, I don’t think AGI/ASI is an impossible idea, but I’m pretty confident that current approaches will not even get us in the ballpark, because they are fundamentally not the right tool for the job. Any allusion to having built the “almost AGI, swear, we’re this close this time” seems, to me, to be little more than marketing hype for silicon valley products and tech stocks. Maybe some day gluing enough of these products together will get you something indiscernible from AGI, but I really do doubt that whole premise. A text transformer won’t become sentient just by throwing more text at it and telling it to process, that’s just a hand-wavy sci-fi premise at best.
The question is whether or not human thought can be represented algorithmically. It seems we agree it’s plausible?
Yea, I think we might agree there but I don’t think that supports the original assertion that human thought is nothing but an (exceedingly complex) algorithm. You can also represent human thought as a system of hydraulic pressures, that’s what early psychology did, and how we got words like repression. But just because you can do that, and maybe even gain some useful knowledge from it - doesn’t mean actual human thought is actually made up of a complex system of pressures/valves - or algorithms. Your map may seem useful, but it ain’t the territory, is what I’m trying to get at, I guess.
To be clear, I don’t think AGI/ASI is an impossible idea, but I’m pretty confident that current approaches will not even get us in the ballpark, because they are fundamentally not the right tool for the job. Any allusion to having built the “almost AGI, swear, we’re this close this time” seems, to me, to be little more than marketing hype for silicon valley products and tech stocks. Maybe some day gluing enough of these products together will get you something indiscernible from AGI, but I really do doubt that whole premise. A text transformer won’t become sentient just by throwing more text at it and telling it to process, that’s just a hand-wavy sci-fi premise at best.