John Hare has an interesting Ares I alternative:

Suppose that instead of stretching and totally redesigning the SRB, they had shifted to a two SRB first stage with the exact same units as used for the last hundred and something shuttle launches. These rockets are fully developed and tested with an extensive flight history of operating in pairs over the last three decades. The purchase costs, handling , and performance are known quantities. Development consists of building an attachment structure, upper stage adapter, and vibration dampening gear. With the considerably more lift performance available from eight segments compared to five in Aries I, the problem fixing payload hits could be absorbed without sacrificing the flavor de jour safety systems NASA would like to have. They wouldn’t even have to game the requirements to match the competition from ULA, Direct, and various upstarts. While it’s possible that this would cost as much as the projected Ares I, it shouldn’t, and if it did, it would be for a system nearly twice as capable.

The first thought, of course, is that it increases the marginal costs quite a bit, but as Jon notes, you can buy a lot of extra SRBs for the billions you might save in trying to tart up the pig. As I noted in comments, one other advantage is that it would allow the first stage to do its own roll control with the SRB gimbals (something that Ares I is incapable of, necessitating a roll-control system for the entire stack on the second stage). But thrust asymmetries could be a big problem. Shuttle can compensate by gimballing the SSMEs, but this configuration wouldn’t have that capability.