lukeprog comments on Justifiable Erroneous Scientific Pessimism - Less Wrong Discussion
You are viewing a comment permalink. View the original post to see all comments and the full post content.
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (116)
This isn't what you asked for, but I might as well enumerate a few of these examples, for everyone's benefit. For the field of AI research:
George Pólya (1954), ch. 15 — a few decades before the probabilistic revolution in AI.
Mortimer Taube (1960) — not long before computers began to regularly dominate amateur and then expert chess players. (Edit: this one seems wrong)
Satosi Watanabe (1974) — a couple decades before both supervised and unsupervised machine learning took off.
Also, Hubert Dreyfus mocked the capabilities of chess computers, and compared AI to alchemy, in Dreyfus (1965) — a mere two years before he was defeated by the chess computer Mac Hack.
Technically, he was correct.
I like the idea of football (soccer) played by quadrupeds.
Taube did not mean "Machines cannot be made to choose good chess moves" (a claim that has, indeed, been amply falsified). Here's a bit more context, from the linked paper.
Taube's point, if I'm not misunderstanding him grossly, is that part of what it means to play a game of chess is (not merely to choose moves repeatedly until the game is over, but) to have something like the same experience as a human player has: seeing the spatial relationships between the pieces, for example. He thinks that's something machines fundamentally cannot do, and that is why he thinks machines cannot play chess.
Now, for the avoidance of doubt, I think he was badly wrong about all that. Someone blind from birth can learn to play chess, and I hope Taube wouldn't really want to say that such a player isn't really playing chess because she isn't having the same visual/spatial experiences as a sighted player. And most likely one day computers (or some other artificially constructed machines) will be having experiences every bit as rich and authentic as humans have. (Taube wrote a book claiming this was impossible. I haven't seen it myself, but from what little I've read about it its arguments were very weak.)
But his main claim about machines here isn't one that's been nicely falsified by later events. We have machines that do a very good job of evaluating positions and choosing moves, but he never claimed that that was impossible. We don't yet have machines that play chess in the very strong sense he's demanding, or even the weaker sense of using anything closely analogous to human visual perception to play. (I suppose you might say that programs using a "bitboard" representation are doing something a little along those lines, but somehow I doubt Taube would have been convinced.)
... Also, Taube wasn't a scientist or a computer expert or a chess expert or even a philosopher. He was a librarian. A librarian is a fine thing to be, but it doesn't confer the kind of expertise that would make it surprising or even very interesting for Taube to have been wrong here.
You accuse lukeprog of being misleading in taking a quote from a mere "librarian", and as we all know, a librarian is a harmless drudge who just shelves books, hence
I accuse you of being highly misleading in at least two ways here:
Mortimer Taube turns out to be the kind of 'librarian' who exemplifies this; the little byline to his letter about "Documentation Incorporated" should have been an indicator that maybe he was more than just a random schoolhouse librarian stamping in kids' books, but because you did not see fit to add any background on what sort of 'librarian' Taube was, I will:
So to summarize: he was a trained philosopher and tech startup co-founder who invented new information technology and handled documentation tasks who was familiar with the cybernetics literature and traveled in the same circles as people like Vannevar Bush.
And you write
!
An upvote for correctly contextualizing what Taube wrote, and a mental downvote for being lazy or deceptive in your final paragraph.
I really can't think of a polite way to say this, so:
Bullshit.
I wasn't accusing Luke of anything; I was disagreeing with him. Disagreement is not accusation. When I want to make an accusation, I will make an accusation, like this one: You have mischaracterized what I wrote, and made totally false insinuations about my opinions and attitudes, and I have to say I'm pretty shocked to see someone as generally excellent as you behaving in such a way.
I do not think, and I did not say, and I had not the slightest intention of implying, that "a librarian is a harmless drudge who just shelves books".
Allow me to remind you how Luke's comment begins. The boldface emphasis is mine.
Taube was, despite his many excellent qualities, not a scientist as that term is generally understood, and he was, despite his many excellent qualities, not working in "the field of AI research".
(Yes, I know the Wikipedia page says he was "a true innovator in the field of science". Reading what it says he did, though, I really can't see that what he did was science. For the avoidance of doubt, and in the probably overoptimistic hope that saying this will stop you pulling the same what-a-snob-this-person-is move as you already did above, I don't think that "not science" is in any way the same sort of thing as "not valuable" or "not important" or "not difficult". What the creators of (say) the Firefox web browser did was important and valuable and difficult, but happens not to be science. What Beethoven did was important and valuable and difficult, but happens not to be science. What Martin Luther King did was important and valuable and difficult, but happens not to be science.)
Pointing this out doesn't mean I think there's anything wrong with being a librarian. When I said "a librarian is a fine thing to be", I meant it. (And, for the avoidance of doubt, it is my opinion both when "librarian" means "someone who shelves books in a library" and when it means "a world-class expert on organizing information in catalogues".)
Now, having said all that, I should add that you are quite right about one thing: when I said that Taube was neither a computer expert nor a philosopher, I was oversimplifying. (Not least because I hadn't looked deeply into Taube's career.) He was an important innovator in the use of punched cards for document indexing, which is quite a bit like being a computer expert; and he was a PhD in philosophy, which is quite a bit like being a philosopher. None the less, I stand by what I said: neither being a world-class expert in document indexing, nor knowing a lot about punched-card reading machinery, nor being a PhD in philosophy, seems to me to be the kind of expertise that makes it particularly startling if one's wrong about whether machines can play chess.
(And, once again, for the avoidance of doubt, I am not in the least trying to belittle his expertise and creativity. I just don't see that they were the kind of expertise and creativity that make it startling for someone to be wrong about the possibilities of computer chess-playing.)
[EDITED to clarify a bit of wording and add some emphasis. ... And again, later, to add a missing negative; oops. Also, while I'm here, two other remarks. 1: I regret the confrontational tone this exchange has taken; but I don't see any way I could have responded sufficiently forcefully to the accusations levelled at me without perpetuating it. 2: I see a lot of downvotes are flying around in this subthread. For the record, I haven't cast any.]
Thankyou for your research. I was mislead by the grandparent.
"Eliezer" should be "lukeprog".
Hah, whups. And so it goes - you correct Eliezer's lack of examples, gjm corrects your description of Taube, I correct gjm's description of Taube, and you correct my description of gjm's description...
Would a chess program that has a table of all the lines on the board that keeps track of whether they are empty or not and that uses that table as part of its move choosing algorithm qualify? If not, I think we might be into qualia territory when it comes to making sense of how exactly a human is recognizing the emptiness of a line and that program isn't.
Yup. I strongly suspect that Taube was in fact "into qualia territory", or something along those lines, when he wrote that.