If by “predict” what Friedman means is “give relatively accurate quantitative magnitudes for certain economic variables in the future” then no Austrian economist has denied that such prediction requires empirical knowledge. As Mises put it, such prediction is based on “thymology” which is a blending together of economic theory (which is universal truths about cause and effect relationships in human action and can be known a priori) and the relevant contingent features of the particular case to be predicted (which can be known only by experience).
If by “predict” what one means is “give accurate qualitative changes for certain economic variables in the future” then theory can be sufficient. One could have accurately predicted the reduction in farm output from the Soviet collectivization of agriculture without knowing anything from experience about such events. In theory we can stipulate particular contingent factors and abstractly deduce what will occur under the stipulation.
What should be asked of Friedman is to name any prediction with meaningful predictive value made by an economist who accepts the empirical-hypothesis testing method that does not require abstraction in the formulation of its model.