Not considering the potential for AI consciousness and suffering seems very shortsighted. There are plausible reasons to believe that both could emerge from an RL processes coupled with small architectural and data regime changes. Today's models have inherent architectural limits around continual learning that make this unlikely, but that will change.