so I guess these days, the fun way to defend the chinese room argument, in an era where computers have damn-near equivalence (not quite, but closer in a way that's a powerful testament to computers being able to "do language"), is to draw a distinction. And say "oh, of course computers can do language! Never said otherwise. It's just different from Formal Symbol Manipulation(tm)."

But I've got numerous problems with this.

Β· Web Β· 1 Β· 0 Β· 0

for one, this is not at all seriously owning the way the argument has been used for decades and decades, where that distinction is almost never observed, and where the issue at stake in virtually every conversation where the chinese room argument is invoked is some big picture argument about what computers are ultimately capable of doing.

Tons of arguments on this front would have collapse if everyone invoked it agreed that computers actually could do language, just that it was complicated.

Show thread

For another, the notion that you can break out "formal symbol processing" from computing generally, as if it's a special subcategory, strikes me as pretty asinine.

It's not just some special subcategory, it's the theoretical basis for any and all computing, including the really intelligent machine learning that is mastering language today.

AI/ML whatever you want to call it has unique approaches, but those rest on top of, and are built out of, the stuff searle labels as "symbol processing"

Show thread

So seeing the progression made in recent years by computers in comprehending language, that seems to be bringing us closer every day to decisively refuting John Searle's entire argument, and responding to these events by saying "oh, no, of course computers can do language! Just not dumb computers" feels frustratingly insincere and revisionist.

If that was really the point the whole time what was at stake in litigating that distinction for the past 60 years.

Show thread

I would also emphasize that there's no natural step from "it's not symbol manipulation" to generative adversarial networks, so I don't feel like any real conceptual foundation was being laid down that lead to or naturally connected the chinese room argument.

It's an awful neat coincidence that invocations of the Chinese Room, so far as I can tell, only really came up to support declarations about the limits of computing, if the real point of it was to highlight other forms of computing.

Show thread

And at the end of the day the substrate of concepts and structures we currently use, to produce something that seems really close to actual understanding of language, a conceptual substrate and programming approach that presumably Searle & co would point to as an alternative to "mere symbol manipulation", is itself a set of structures built up out of a foundation that is symbol manipulation. It's not a different kind of computing. It's not a different theoretical foundation.

Show thread
Sign in to participate in the conversation
ACP πŸŽ‰πŸ°

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!