I mostly agree with you, but I suspect that a young AGI growing up in a community of humans could still develop a fully-functional self-model and ascend the Maslow hierarchy... I don't see why an AGI society would be required...
-- Ben G On Fri, May 3, 2013 at 3:19 PM, Chris Nolan <[email protected]>wrote: > >Our plan with OpenCog is to embody it in video game agents and mobile > robotic agents, so that it would be an "agent" from the start.... > > Maybe I should separate out what I mean as a difference between a simple > agent and one that is a complex, self-referential agent. A complex, > self-referential agent would be one that is operates within a complex > environment, understands itself as an actual entity within that > environment, and seeks fulfillment of the upper levels of the Maslow > hierarchy of > needs (belonging, self-esteem, self-actualization). What I mean by a > "simple agent" would be one that is able to function in an environment in > autonomous way but does not necessarily view itself as as existing as an > individual creature. It might have a high-level reasoning capacity and even > operate within a complex environment, but does it go beyond the lower-level > hierarchy of needs towards the upper "human-level" needs, and possibly > beyond? > > I think that is how I would separate out between sub human-level AGI and > full human-level AGI and beyond. My intuition in that regard is it will > require the evolution from lower level to higher levels within a complex > Multi-Agent System; i.e. co-evolving an AGI society along with a higher > levels of AGI. The reason I have that intuition is that I think the complex > nature of "self-identity" for humans is intrinsically tied into our > interactions with society, even if at a subconscious level, and can't be > separated out from our growth in society from childhood to adulthood. The > dynamics of that interaction wouldn't be that much different for AGI. The > complexity of that is such that it couldn't be programmed but would instead > have to be evolved and grown. Maybe that would be virtual, maybe that would > be the interaction of many robots in human culture, I don't know. But I > think it would require that interaction to go from sub human-level to full > and post human-level AGI. > > Maybe I'm wrong though. As I said, just my intuition... > > > ------------------------------ > *From:* Ben Goertzel <[email protected]> > *To:* AGI <[email protected]> > *Sent:* Friday, May 3, 2013 12:23 AM > > *Subject:* Re: [agi] Toward enlightened AGIs > > > An AGI society would be awesome... however, I doubt it's critical... > > Our plan with OpenCog is to embody it in video game agents and mobile > robotic agents, so that it would be an "agent" from the start.... But > various OpenCog components are now being used as tools... > > -- Ben G > > On Fri, May 3, 2013 at 8:14 AM, Chris Nolan <[email protected]>wrote: > > Ben, > > >I think that humans have an evolved tendency to be overly egocentric, and > (esp. in modern Western culture) to model themselves as isolated, separate > beings much >more than is really the case.... So compassion meditation is > in part a way of overcoming this particular human propensity.... OTOH, AGI > systems would not >necessarily have that sort of propensity in the first > place... > > I don't necessarily disagree. A slight difference I would state is that > humans have evolved the natural ability to feel compassion, however it's > often linked to group identity; i.e. we naturally feel compassion for > family and those we consider members of our "tribe" but not necessarily for > those outside. Compassion meditation is then a way to change that neural > wiring to create that natural compassion for those whom we might normally > consider "outside our tribe," or even for the "enemies of our tribe." AGI > systems might not have the bias to tribal identity and creating the > "other," but I wonder if the natural bias might be towards creating no > group identity at all and so the unstated bias would then be for all to be > "other"? > > >We definitely would, however, want our AGIs to have an initial bias to > model others and see the world from other beings' views.... This militates > toward a kind of >non-attachment to one's own > self/self-model/individual-perspective... > > Yeah, I think that's a very good point. There's a very big difference > between non-attachment to self because of a viewpoint of being connected > with a larger whole and non-attachment to, or really detachment from, self > because of a lack of connection to any group or identity. How do you see > this being implemented within an AGI system? Would it be necessary to > have it explicitly designed within OpenCog's framework or would it be a > part of the implicationLinks that are learned through experience? > > This also brings up another question which is what role will the creation > of multi-agent systems play in the evolution of AGI? It seems likely to me > that either OpenCog or some other AGI system will eventually create a > general AI reasoning capacity, an AGI tool, that is human-level. But, will > that be the same as the creation of an AGI agent, or will it be necessary > to then evolve the AGI from a tool-level to an agent-level within some > larger multi-agent system (likely a combination of humans and other AGIs). > My intuition would be yes because I think that is how you would evolve the > AGI's understanding of itself within the larger group context, creating the > experience and implied understanding of behavior within the larger group > (which then necessitates a model of self). The next question begged is then > what size magnitude of multi-agent system? Would only a few AGI agents > interacting with humans be enough, or would it be more beneficial to have a > very large number of AGI agents (an AGI society)? Or, stated another way, > is it necessary to co-evolve AGI society with the AGI agent? > > It is a part of how human intelligence has evolved... > > ------------------------------ > *From:* Ben Goertzel <[email protected]> > *To:* AGI <[email protected]> > *Sent:* Thursday, May 2, 2013 3:38 AM > *Subject:* Re: [agi] Toward enlightened AGIs > > > Interesting response, thanks... > > I think that humans have an evolved tendency to be overly egocentric, and > (esp. in modern Western culture) to model themselves as isolated, separate > beings much more than is really the case.... So compassion meditation is > in part a way of overcoming this particular human propensity.... OTOH, AGI > systems would not necessarily have that sort of propensity in the first > place... > > We definitely would, however, want our AGIs to have an initial bias to > model others and see the world from other beings' views.... This militates > toward a kind of non-attachment to one's own > self/self-model/individual-perspective... > > ben > > > > On Thu, May 2, 2013 at 8:39 AM, Chris Nolan <[email protected]>wrote: > > Ben, > > that's an interesting concept. From my reading of Buddhism is also seems > like non-attachment in meditation is also often linked with metta practice, > or compassion meditation to state it in a simplified way. Have you looked > at any of the neuroscience papers on that practice? In the simple example > you supply of "Bob" detaching from his girlfriend the practice would be not > just "letting go of the suffering of the break-up" but also adding a > compassion practice for ex-girlfriend, in this way the Buddhist practice > would be developing non-attachment in conjunction with compassion for the > individual and their choice. In this way an individual avoids just falling > into the trap of avoiding suffering and so getting caught by it even more. > Side note: as someone having been in a number of break-ups I've found it > works better than just trying to detach, haha... > > I bring it up because I wonder if the concept could be informative for the > goal of creating a Friendly AI? In this way OpenCog's system of balancing > attachment and experience could also be linked with broader compassion. > Possibly in the implication links, while disassociating happiness with > "put_arm_around_girlfriend" > and adding an implication that happiness for girlfriend includes > separation from Bob. That possibly hints at way of formulating ethics for > A.I. > > -Chris > > > ------------------------------ > *From:* Ben Goertzel <[email protected]> > *To:* AGI <[email protected]> > *Sent:* Wednesday, May 1, 2013 1:24 AM > *Subject:* [agi] Toward enlightened AGIs > > For your general amusement, here is a blog post I rote on > > "The dynamics of attachment and non-attachment in human and AGI minds": > > > http://multiverseaccordingtoben.blogspot.hk/2013/05/the-dynamics-of-attachment-and-non.html > > :) > ben > > > -- > Ben Goertzel, PhD > http://goertzel.org > > "My humanity is a constant self-overcoming" -- Friedrich Nietzsche > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/20347893-f72b365c > Modify Your Subscription: https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com > > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/212726-deec6279> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com/> > > > > > -- > Ben Goertzel, PhD > http://goertzel.org > > "My humanity is a constant self-overcoming" -- Friedrich Nietzsche > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/20347893-f72b365c> | > Modify <https://www.listbox.com/member/?&;> Your Subscription > <http://www.listbox.com/> > > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/212726-deec6279> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com/> > > > > > -- > Ben Goertzel, PhD > http://goertzel.org > > "My humanity is a constant self-overcoming" -- Friedrich Nietzsche > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/20347893-f72b365c> | > Modify <https://www.listbox.com/member/?&;> Your Subscription > <http://www.listbox.com/> > > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/212726-deec6279> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com> > -- Ben Goertzel, PhD http://goertzel.org "My humanity is a constant self-overcoming" -- Friedrich Nietzsche ------------------------------------------- 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
