Lasers were thought to be a laboratory curiosity when they were invented. On the other hand, you have a ton of academics who are convinced that the "semantic web" is the next big thing.

Part of the problem with science is that it's an activity where it's very hard to have any useful metric. Even experts in the field have a tough time deciding whether a given research direction is going to be fruitful. They're often wrong.

So far the best thing we've figured out for science resembles the Underpants Gnomes' business plan:

1) Fund lots of science.

2) ????

3) Profit!

The ???? part is "pray that some of it is useful and that some of that gets into the hands of people who will use it."