Humans are genes' way of reproducing themselves. Human behavior implements a utility function all of whose terminal values are statements about gene allele frequencies. As humans were not even aware of allele frequencies until recently, anything a human thinks it values cannot be a terminal value of a human utility function.

I don't find this obvious. Genes may have the goal of reproducing themselves and certain genes have adopted a strategy of using humans to do so. If we model human behavior by a utility function, it makes sense to assert that this would depend on the genes--if it didn't, it's unlikely humans would have evolved in the first place. So far, so good.

The question is whether the interactions of ~20,000 genes among themselves and with the environment doesn't add anything extra. Figure that genes are really just elementary particles arranged in a certain configuration, yet the gene's quest for reproduction does not have a clear origin from the elementary particles themselves. It's worth asking whether scaling up from genes to humans has its own emergent phenomena, where a human has values that don't obviously correspond to those of the genes.