I guess that developing learning capabilities within VR gives the ability to repeat the same experience indefinitely, the real world doesn't work that way. You can let a learning program playing a game 1000 times during the night and the following day it has leaned how to play it. It is good to speed up the testing of a learning algorithm. However, at a certain point those algorithms will have to face the real world. How much VR training is necessary to be able to start operating in the real world is not known, each Agent will develop differently. This doesn't mean that there shouldn't be some smart ways to bypass the VR phase altogether.
On Sun, Oct 25, 2015 at 9:20 PM, Korrelan AI <korrelan...@gmail.com> wrote: > I tried using a physics engine and virtual world but there is no > substitute for the real world. > > > > I wrote audio/ visual/ tactile etc input modules than can be run on > separate physical processors/ machines. The main software is designed to > be parallel and so can also be run on multiple machines. A 1GB backbone > network for the sensory streams and inter column communications and a ram > disk on the master machine for flagging/ coordinating the rest of the > Beowulf cluster. > > > > Crude but effective, gets around the whole virtual world problem. > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/27738727-35a82580> | > Modify > <https://www.listbox.com/member/?&;> > Your Subscription <http://www.listbox.com> > ------------------------------------------- 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
