I would say his point is more that classical GR actually follows (holographically) from quantum mechanics (without gravity) in a suitable limit. For example, if the quantum mechanics model is some strongly-interacting CFT, you expect the dual theory to be well-described by classical GR physics. He then proposes to simulate a CFT with quantum computers to effectively "test" the corresponding gravity theory. He also goes beyond classical GR. He points out that the Newton constant G in the dual theory is not necessarily small (it depends on the CFT), so you could actually probe strong-coupling quantum gravity dynamics with such a simulation. Like many high-energy theorists, I think he overestimates how close we are to doing these simulations on quantum computers.