I don’t think AGI is necessarily impossible, but I’m not convinced that it’s possible to achieve in a way that gets around the constraints of human intelligence

I don’t think AGI is necessarily impossible, but I’m not convinced that it’s possible to achieve in a way that gets around the constraints of human intelligence. The singularity idea is basically the assumption that AGI will scale the same way computers have scaled over the past several decades, but if AGI turns out to require special hardware and years of training the same way we do, it’s not obvious that it’s going to be anything more remarkable than we are.