CMU's Aran Nayebi argues model introspection is better evidence for AI interiority than matching global workspace architectures
He notes human neurological data poorly supports the global workspace theory.
Story Brief
He notes human neurological data poorly supports the global workspace theory.
Source Posts on X
Cluster sources
Aran Nayebi@aran_nayebiTECH
Right, it's about the former, rather than the latter (and it's worth noting there isn't strong evidence for global workspaces in human data anyhow: https://x.com/aran_nayebi/status/2074221765811744865). I think the earlier introspection results more convincingly establish interiority, though it's worth exploring further.
David Manheim@davidmanheimTECH