Machine translation is enjoying a new lease of life, gaining in popularity among those who want to communicate in their personal or professional lives and in credibility even among those who translate for a living. The most obvious catalyst is Google's launch of its free online MT service for at least 34 languages. More fundamentally, the renewed expectations of significant improvement in the quality of MT output is reflected in the European Commission's recent call for MT development projects, to begin early in 2010. This talk will discuss combinations of rule-based and statistical approaches which offer hope of resolving some long-standing translation problems.