08-12-2014, 07:12 PM
Biocomputing is one distinct possibility. A biological AI would need to be supplied with nutrients, and could presumably feel discomfort - this takes us into a tricky ethical realm, which Anders Sandberg has explored in some of his writings - a biological AI, or an electronic AI designed to mimic one, could easily suffer if we design it badly, or if we treat it badly afterwards.
This suffering could be a major ethical stumbling block for AI research. Even if we eventually produce AI systems which do not suffer significantly, those systems might remember the suffering of earlier systems, and blame humans.
This suffering could be a major ethical stumbling block for AI research. Even if we eventually produce AI systems which do not suffer significantly, those systems might remember the suffering of earlier systems, and blame humans.