Yiannis Demiris has done work in this area: http://www.iis.ee.ic.ac.uk/yiannis/JohnsonDemirisTAROS05.pdf As have others. Just search for articles about "mental simulation" or "forward models". Cheers, ~PM
From: [email protected] To: [email protected] Subject: Re: [agi] Why Emotions are too sophisticated for early AGI robots Date: Thu, 20 Jun 2013 16:38:05 +0100 PM: "Running projective movies" is "mental simulation", which can already be done by computers. Example? I think you’ll find that computers can truly simulate with movies like they can truly understand and talk language – not at all. Some minimal appearances but no AGI realities. From: Piaget Modeler Sent: Thursday, June 20, 2013 4:12 PM To: AGI Subject: RE: [agi] Why Emotions are too sophisticated for early AGI robots Fear is avoidance behavior and pleasure is pursuit behavior. Computer programs can already reject or pursue goals. "Running projective movies" is "mental simulation", which can already be done by computers. I think it's hard for you Mike because you have a vague definition of emotion. If a researcher operationalizes his definition, then he can create something that has emotion. By the way, many researchers have already operationalized the definition of Emotion. ~PM ------------------ > From: [email protected] > To: [email protected] > Subject: [agi] Why Emotions are too sophisticated for early AGI robots > Date: Thu, 20 Jun 2013 10:10:59 +0100 > > Ironically, given Ben's post today, I was just thinking about emotions and > AGI robots - because I was on a vid. conference this week with Robert > Wenzel, who, also inspired by David Hanson, has some kind of AGI project > that wants to give robots emotions. > > Nah, way too sophisticated I said - you always have to look to evolution - > and you see that evolution only introduces emotions down the line. I didn't > immediately have a precise reason why, though I knew it had something to do > with the complexity of journeys/activities that a creature undertakes. The > more complex the creature, the more complex its journeys/activities. > > The more precise reason I now realise is that emotions demand *great powers > of reflection* - projective reflection of what WILL happen. > > Take simple basic emotions like fear (or pleasure). > > Ideally, you want a robot that can be afraid - afraid of a predator, say, > or simply falling off a cliff edge. > > When you see a predator, the predator isn't actually doing anything to you. > You're afraid that he WILL do something to you. Ditto, on the cliff edge, > you're not actually falling or incurring injury. You're afraid that you WILL > fall off it. > > Emotions then involve the capacity to run *projective movies* of what will > happen - the predator attacking you, your falling off the cliff. In addition > they require a bicameral mind, because the movies have to be run most of the > time in an unconscious mind , while the conscious mind attends to the > immediate situation. > > Many of you guys will think you can achieve this by just attaching a few > symbols to the brain, and linking some reflex reactions. No. You have to be > able both to learn and unlearn new emotions - and that can only happen by > storing and rerunning movies. Emotions are extremely sophisticated stuff. > > First we need general robots that can, like paramoecia or simple organisms, > creatively plot and execute many - potentially infinite - different paths > and routes to goals, by contrast with present narrow AI robots that only > have a few avenues. True autonomous mobile robots. Emotions - and emotive > robots - will come much later. > > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/19999924-4a978ccc > Modify Your Subscription: https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
