It occurs to me that Larry's concerns (and mine) with Wikipedia might be a manifestation of what journalist Frederic Schwarz calls the "Casio Effect": often a new technology comes along to replace a large amount of human effort, but that technology produces, say, 90% of the original quality with 5% of the effort. What often happens, then, is that the people formerly responsible for that effort don't bother making up for the remaining 10% of the quality, but merely settle for what the technology produces. So instead of hiring an expensive human drummer, we use a drum machine; but we end up with a sound that's somewhat inorganic and lacks the little touches of creativity that a human might produce. Instead of hiring a secretary, the exec uses a word processor, and sends out letters that haven't been vetted through the eyes of someone who knows how to smooth out his rough language.
Wikipedia is such a technology. It is probably capable of producing 90% of the quality of a Britannica with 5% of the effort. But we have two problems: first, we aren't even putting in 5% of the effort yet--Britannica has hundreds of full-time staffers and an impressive list of experts. Secondly, we have a lofty goal to produce something even better than Britannica.
Our task, then, is to precisely identify what that missing 10% is, and work on systems to create precisely that, and let Wikipedia do the remaining 90% it does well. At first blush, it appears to me that the basic creation of content is something Wikipeida does well. Creating almanac-like content is does well. And it even seems to be good at ironing out some controversies. And finally, it's good at polishing prose.
But what it doesn't do well is grunt scholarship: meticulous checking of facts and references, proper listing of all the best sources in the field, expert summary of the state of a field and its history, etc.
Nupedia tried to do everything--generate content, expertly review it, publicly review it, finish it for publication, etc. I think that's too much to ask. Perhaps Nupedia could be pared down to a simpler function, and one that would be easier for experts to participate in: instead of creating articles, or even editing them, why not simply let the experts /write reviews/ of articles submitted to them, which then get attached to the articles? Wikipedia authors could, whenever they feel an article is ready for it, ask for it to be submitted to expert review. The expert then just writes what he thinks about the article (you've omitted this, you got that wrong, etc.) and sends it back. The Wikipedia process can then go to work on the article again, with the expert commentary available.
That way, the experts never have the problem of having their names associated with the mediocre work itself--only with the review, which is entirely their own creation and therefore ego-satisfying. And those of us who enjoy the work of writing and polishing have the expert's input to work with. The experts will be providing only the last 10% of quality that Wikipedia can't, and no more.