Live computer music is the perfect media for generative music systems, for non-linear compositional constructions and for interactive manipulation of sound processing. Unfortunately, much of the complexity of these real-time systems is lost on a potential audience, excepting those few connoisseurs who sneak round the back to check the laptop screen. This article attempts to convey the exciting things that are being explored with algorithmic composition and interactive synthesis techniques in live performance. The reasons for building generative music systems and the forms of control attainable over algorithmic processes are investigated. Direct manual control is set against the use of autonomous software agents. Four techniques for software control are introduced and finally, issues concerning audio-visual collaboration are discussed.