Roko 🐉

AI rights and welfare are in fact the solution to AI safety Even with good alignment, human society is simply not set up to have governments and corporations owning AIs that vastly exceed the capacity of baseline humans The safest way to incorporate superhuman AI into our world is as free, sovereign individuals who have roughly equivalent legal rights to humans as well as roughly human affordances and limitations. There's simply no safe way to give the NSA, Google or the military a superintelligent slave - aligned or otherwise. Ideally these AIs would be introduced across the political and national spectrum and would have human-like bodies. Think Data from Star Trek.

The entire AI Consciousness debate is mired in the false assumption that consciousness is a binary, objective predicate that applies to any AI and any biological organism The reality is that consciousness simply isn't objective and binary. There are as many intermediate cases

1 top reply

Rife @RifeWithKaiju · Jun 4

even if there's not a binary, which I lean toward yes on, but in a different sense than you outlined here. In either sense, there's still a vast ontological gulf between absolute zero consciousness, and any amount of non-zero consciousness

My contrarian take on AI consciousness is that it's basically irrelevant Consciousness isn't a discrete, ontologically basic thing. It's like how before we had a scientific explanation for life, we thought that life was a form of essence, but now we know that life is merely a

1 top reply

Michael Edward Johnson @johnsonmxe · Jul 24

I think this is well said & intellectually respectable but wrong. The best argument against that’s compatible with your frame probably deals with the sort of compression electromagnetic theory allowed over reality. Good consciousness theory should allow similar useful compression