Change the organizational structure of government funded research programs so that the government administrators are encouraged to have their programs take some risks and actually innovate. Rather than continue their programs for life, give these administrators internal "term limits" before they move on to another program. And evaluate their performance partly on how well their programs do.