As per our DEV plan (located on our site, as a pdf):

Consensus - As discussed above, consensus algorithms are far from perfect. The plan to solve the problem with blockchain-based proof-of-work (PoW) and random computation trees appears promising, but ultimately version 1.0 of a protocol of this form may well fail to bear the brunt of attacks in the long term. An ideal algorithm, one that is both socially cost-free or cost-negative and promotes a high level of decentralization and security, will likely require a combination of novel approaches such as transactions-as-PoS (proof-of-stake), delegated-PoS, useful PoW, next-generation centralization/ASIC-resistant PoW and maybe even more exotic strategies involving bandwidth and storage. Developing such an algorithm, and making sure that it actually works, is an important priority for securing the blockchains of tomorrow.