P.S. It is also possible to use the "Self-Indication Assumption" as an alternative to the "Self-Sampling Assumption". Or to use a non-anthropic model like "Full Non-Indexical Conditioning". But these don't get rid of the argument that we are unlikely to turn into a post-human society, and for some rather interesting reasons, which Katja Grace discusses here

So far, I think reference classes are the only counter-argument that might work.