an activity that would be very handy to be able to do, but is not (to my knowlege) currently implemented. The main reasons for this (that I can think of on the spot) are that definitions are hazy, and involve natural language.

For instance, if I were to ask the average person, "what's the name of that fuzzy white stuff that grows on bread"1, the reply would (should) come back almost instantaneously, "mold." But if I were alone, with noone around to ask, how could I find out? I suppose that I could browse through a dictionary until I happened upon a definition like the one I was looking for, but "m" is a long was from "a". Creating a reverse-lookup dictionary would not be feasible, because unless you included every permutation of every definition, like "smooth white stuff", or "fuzzy green stuff", instead of "fuzzy white stuff", it would still be next to impossible to find anything in it.

All that said, it would be very cool to be able to ask some computer program (or Everything), have it work some magic, and have it spit out a word that is a pretty good match for your definition. Perhaps Latent Semantic Indexing, or some computational linguistics wizardry; perhaps not...

1 dscotese points out that this is not the best example sentence; if you search for it using Google, the first link is to a page about mold. On the other hand, "fuzzy white and sometimes green stuff that grows on bread", as of Dec 19 2003, returns something about a girl in Austin with a fuzzy black sweater, so I think my point still holds.