How well is the modern-day starburst-AGN connection mirrored in the early Universe? This is starting to be answered by deep wide radio surveys such as ATLAS, which are giving us a new view of high redshift galaxies. For example, we find powerful radi
o-loud AGNs which look like star-forming spirals in the optical and infrared, a composite which is almost unknown in the modern Universe. We find radio-bright objects which are unexpectedly invisible in the infrared, and which may be very high redshift radio galaxies and quasars. And although the radio-far-infrared correlation for star-forming galaxies has now been extended down to microJy levels, we still cannot reliably distinguish between starburst and AGN. So what do we need to do to ensure that SKA and its pathfinders will be able to understand galaxy evolution in the early Universe?
