The only problems would be speed and memory.

There is a tiny chance that when he said "does not reproduce the causal structure of neural interactions", what he actually meant was "would simulate the neural interactions extremely slowly", but if that was the case, he really could have said it better.

My priors are that when people without formal computer science education talk about brains and computers, they usually believe that parallelism is the magical power that gives you much more than merely an increase in speed.

In practice it's just a matter of computational power. His statement makes it fairly clear that he doesn't understand this distinction.

Circuit level simulations of advanced microchips certainly exist - this is not just theory. Yes they are super expensive when run on standard CPUs (real-time simulation of an iphone CPU naively would require on the order of an exaflop). However, low level circuit binary logic ops are much simpler than the 32/64 bit ops that CPUs implement, and there are more advanced simulation algorithms. Companies such as Cadence provide general purpose binary logic emulators that actually work, in practice for reasonable cost, not just theory.