Of course, it would be cool if Bitcoin didn't require throwaway computation. Then we could use this computation for something that's actually valuable, like protein folding, or something, and still make use of the waste heat.
That "throwaway" computation ensures the integrity & correctness of the blockchain, all while confirming new transactions.
Other proof of work algorithms could concurrently secure the block chain and also provide direct, external benefit. There are some suggestions elsewhere in this thread. :)