> Didn't the same argument apply in general to why NNs weren't particularly useful for anything ~30 years ago?

Not unless there was good reason to expect that every other possible application of NNs would be as difficult to achieve as general intelligence.