« links for 2007-05-06 | Main | links for 2007-05-10 »

Hybrid methods, then

"Unifying Reasoning and Search to Web Scale".

Why do AI types keep having to discover hybrid methods and/or resource bounded agents?


May 8, 2007 11:18 PM

Comments

Bijan Parsia
(May 9, 2007 08:08 AM #)

First, I don't get anything for the first link.

Second, I'm really curious as to the people you intend to tag with your description "AI Types". I, for example, and most people I know who work in the field, don't have to discover them (though, uhm, all agents are resource-bound; one way to deal with time-space limitations in the face of, e.g., NEXPTIME problems is to optimize; another is to approximate).

Third, ooops, I guess you managed to troll me. Good show!

Bill de hOra
(May 9, 2007 08:13 PM #)

First, the first link gave me something.

Second, perhaps "GOFAI" types is sufficient to give you a clue. It's quite bizarre that the semweb community collectively went down a representation is everything black hole for half a decade almost ensuring non-adoption. What was that about? Tenure? Uhm, at least people are coming around to the idea that representation isn't everything, or even where to start.

Third, it's no troll. The point is that semweb research has played out much like AI research. Oops!

Bijan Parsia
(May 10, 2007 10:39 AM #)

Hmm. I'm still not getting anything for the first link in Safari. Ah, Firefox does the trick.

No, it's not. I don't believe that "AI types" or "GOFAI types" represent any recognizable grouping in semweb or ai research today. (Care to provide at least departmental level indication of whom you mean?)There are people working on first order logics and variants. There are people working on non-monotonic logic. There are people working on probabilistic systems, etc. These are all fairly robust fields.

I don't know anyone *in academia* (thinking again...nope) who ever advocated "representation is everything". I don't even quite know what that *means*. I certainly don't see how caring about KR is in conflict with the page you link to. (Note that the KR conferences is really KR *&* R.)

It is, of course, always convenient to blame something for "non-adoption" and then to blame that something on a class of people that are the traditional goat in this area. Oh, and your alternative guiding principle is also not contentful (and to the degree it has content...i.e., the accusation that (even a plurality of) people (doing research on the Semantic Web) thought representation is everything...is just false. *Show* me that this is true by even the slimmest and most preliminary of empirical evidence. If you can do that, then we move to a non-troll situation.

So, it is a troll. "Foolish AI people have fucked it up again" is a more than fair paraphrase of your post. Since "they" arguably didn't fuck "it" up the *first time*, thus, you've trolled.

Bijan Parsia
(May 10, 2007 11:38 AM #)

(Btw, I *think* I'm taking into account hyperbole. I.e., I'm not trying to hold you to the standard of someone having to have literally said, "Representation is everything". Conversely, I expect that things like Pat Hayes' *Naive Physics Manifesto*, which advocates, among other things, worrying about representation "first", for, say, a generation not to count against KR types. After all, it's not like a dominant number of people stopped taking the constraints of reasoning and of application needs into account subsequent to NPM.

In other words, I am trying to give you the benefit of doubt when possible.)

Post a comment

(you may use HTML tags for style)




Remember Me?

Trackback Pings

TrackBack URL for this entry:
http://www.dehora.net/mt/mt-tb.cgi/2086