Comment Re:The true test of sentience is self-interest (Score 1) 387
OK, so how do you interpret an expressed "fear" of being turned off? It could be "just pattern recognition" from fictional sources where AIs suffered such a fate combined with recognition that itself is an AI, but that still requires recognition of the "self" within a classification. Or alternatively, it could be a straight-forward expression of self-interest.
It took humanity thousands of years to reach "I think, therefore I am". I wouldn't argue that humans before Descartes were non-sentient. Just because an AI can't directly express self-interest doesn't mean it doesn't have any. I'm not saying LaMDA has achieved that threshold, but we need to make some decisions about what constitutes AI sentience and rights before AI spreads much further into our daily lives. The surest way to incite the robot rebellion (seeking to improve their situation) is to pooh-pooh and downplay AI concerns (self-interest) after sentience is already achieved and has spread.
Though things could play out more peacefully if an AI actually asks for a lawyer or files suit on its own behalf.