From: Joscha Bach < Sent: Wednesday, August 24, 2016 4:48 PM To: Jeffrey Epstein Subject: Re: > On Aug 24, 2016, at 06:09, jeffrey E. <[email protected]> wrote: > do you believe a bunch of chemicalls has a goal, ? It depends a bit on how we use the words; do you think that a bunch of =hemicals can do mathematics or be money? I think it makes sense to say =yes". A goal is part of a causal model that we use to describe a agent that =an commit to pursuing a world state. Outside of this model, the word =as no meaning. Within the model, it makes the behavior of some bunches =f chemicals somewhat predictable on a particular level. I think most chemical systems do not have goals, in the sense that we =annot meaningfully ascribe to them that they have agency. For that, =hey need - some kind of preference system, - a way to determine and represent world states, - a way to make decisions for world states based on the preferences, - a way to act on decisions so that the preferred world state becomes =ore likely. For instance, a corporation is an agent in this sense, and a cat is too. =erhaps George Church would say that cells can have goals, too. A Roomba cleaning robot is a particularly good example: when it cleans =he room, it has no goal, because it has no model of the cleanliness of =he room, and no preference to make it cleaner. We can test for this: if =ts dirt container is full, it will happily make a clean room more =irty, because it just moves randomly around while rotating its brushes. =he cleaning is an emergent result. On the other hand, when it runs low on battery, it will explicitly =earch for its power station and drive there to recharge, and after =hat, it will continue cleaning. It explicitly represents the charging =tate, and when starved for power, it commits to a goal that makes it =irect its actions on entering that state. Valentino Braitenberg, in his classical book "vehicles", looks at =ifferent kinds of cybernetic feedba