Hi Chris — Yes, sparing partners is a good way to think about it. And with multiple machine collaborators, some will be more supportive, and others can be more adversarial, as well as just riffing off of our ideas.
Your random experiments are great as well, and I’m finding that randomness and a whole range of other fairly simple techniques are great additions to actual machine learning. This all fits in with the “dumb-smart” approach, and definitely draws on approaches like Oblique Strategies.
I’ll be sharing some experiments and thoughts about the nuts and bolts of crafting these personalities soon.