Report Blog Entry
The Chinese Room is a shit argument and Searle has done more harm to the AI industry 2nd only to the Perceptrons book.
Instead consider the Mandarin room. Same setup. Slips of paper. Man in the room, just following instructions. But instead of a filing cabinet, there's a small child from Guangdong in there that reads it and tells the man what marks to make on the paper he hands out. oooooo, aaaaaaah, does the man know Mandarin or doesn't he?!? Shock, gasp, let's debate this for 40 years! Who cares what the man does or doesn't know. And talking about the room as a whole is a pointless waste of philosophical drivel.
It's a 3-card monty shuffle. If you had a book or filing cabinets that were big enough and followed it's if-this-then-respond-with-that logic that could hold an unbounded conversation the book would have more pages than could fit in the known universe. There's the magic trick. That's where the intelligence lies. Not in the middle-man, nor the room on the whole. Why is Searle even talking about the man at all? The book is what obviously knows Mandarin. Same too, it's the software that possesses intelligence and understanding and knowing. Humans are just egotistical little shits that want to be special and will dance around this idea looking for a soul to the end of time.
If we were being generous Searle was trying to make an analogy to comment about how components of the whole can't individually know anything. Your RAM doesn't know how to run doom. Your south-bridge doesn't have morality. The CPU just follows instructions. Sure. But in much the same way your spinal cord does not appreciate Mozart nor does your hypothalmus have a moral system nor does any neuron truly know anything. This flows into the arguments about "the room on the whole" and emergent properties. But that's just part of the 3-card monty, when one glossed over component is obviously where the intelligence lies.Consider an EXTREMELY bounded conversation where the user is only allowed to submit ONE letter. And the book can only ever respond with ONE letter. The book tries to form words as best it can. The user can submit whatever they want. The user's "C" can be followed by the book's "A", which could form "CAT". 10 slips of paper going back and forth. Forget any randomness or creativity, it's just a pile of if-this-then-that lines of instructions the man can follow.
With 26 letters (in English), the book needs 26 case statements of one form or another. Let's say we have a clever system that lets the book convey all 26 cases and what to respond with and what page and line number to then turn to all within one line of the book. With 1 exchange, the book only needs one line. It can solve EVERY two letter word! Success! ...But for 3 exchanges, all 26 cases need their own 26 cases to try for another word, and so the book would need 676 lines and would be about 10 pages. This is just basic math. With 10 exchanges going back and forth, the book needs to be 2.13e+12 pages. 2,138,895,388,690. That book is 192,901 MILES thick and wraps around the planet about 10 times.
Searle is offhandedly supposing such a book could hold instructions to have such a back and forth conversation with not only single letters, but whole words and sentences and paragraphs. It's a 3-card monty game of misdirection. Even just stating that such a thing could exist within a room is part of the ruse.
Comments
Displaying 0 of 0 comments ( View all | Add Comment )