[ExI] Watson on NOVA
rpwl at lightlink.com
Sun Feb 13 16:39:25 UTC 2011
Kelly Anderson wrote:
>> The good stuff is between about 15 minutes and 28 minutes. We will have
>> practical companion computers very soon. All doubts I once suffered have
>> vanished with this NOVA episode.
> While I am clearly jazzed about Watson, and I do know for sure now
> that Watson uses statistical learning algorithms, I am not quite as
> convinced that there is a general solution here. At least not quite
> yet. The types of answers generated seemed to have been heavily
> "tweaked" for Jeopardy. That's not to say that Watson isn't
> interesting, and an important milestone in AI. I think it is both.
> Just that it isn't quite as far down the road of machine understanding
> as I had hoped. Some of the video seemed to indicate that it used some
> kind of statistical proximity based text search engine, rather than
> parsing and understanding English sentences quite so much as I thought
> maybe it did.
Sadly, this only confirms the deeply skeptical response that I gave
I strongly suspected that it was using some kind of statistical
"proximity" algorithms to get the answers. And in that case, we are
talking about zero advancement of AI.
Back in 1991 I remember having discussions about that kind of research
with someone who thought it was fabulous. I argued that it was a dead end.
If people are still using it to do exactly the same kinds of task they
did then, can you see what I mean when I say that this is a complete
waste of time? It is even worse than I suspected.
Of course, since NOVA was presenting things on a general
> audience basis, it may have downplayed any NLP aspect.
> This will be useful technology (assuming it escapes research) I can
> see it answering really useful questions. I hope they build it into a
> search engine. But it does, for the present, seem to be very tweaked
> for Jeopardy... which is, I suppose, what I should have expected.
> Has anybody seen any technical papers by the Watson team? That would
> be interesting in evaluating just how they did it.
> Since Watson is essentially a bunch of PCs, I can see this being
> deployed into the cloud pretty easily. And if Watson can look on the
> Internet, then perhaps it can come up with better answers (albeit
> perhaps more slowly) than in the isolated Jeopardy case. It seemed
> that they stuck with Wikipedia, online encyclopedias, the internet
> movie database and other specific information sites, rather than
> crawling the entire web. Perhaps they did this to ensure greater
> accuracy??? Or maybe it was a storage space issue. In any case, if
> they make a bigger machine in the cloud that accesses the internet and
> has more storage, I'm sure they could come up with some very
> interesting answers to general questions, assuming the answers are out
> there somewhere.
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
More information about the extropy-chat