Abstract: We will review the underlying theory for how turbulence makes sound and use that to motivate detailed simulation-based investigation of jet noise mechanisms.  High-fidelity sub-sonic noise prediction is particularly elusive because the wavenumber-frequency makeup of the turbulence is such that it does not directly couple with propagating wave equation solutions.  This will be illustrated with a model two-dimensional mixing layer, which is perturbed into a quiet state based upon an adjoint-based optimization procedure.  The potential for extension of this approach to engineering applications is discussed, including preliminary examples.  Higher-speed jet noise is, in a sense, simpler because is this case the turbulence directly couples with the sound.  However, the radiation is so strong that its generation and propagation involves important nonlinearity, which complicates its description.  Recent high-speed-flow simulations will show some of the key features of these mechanisms.