Besides which, consciousness and agency are two different things, damnit! I've got a github repo in which I train agents by policy optimization. I have no illusions that my Pytorch models are conscious.