There are numerous ways in which a goal could be achieved, and in the human case this is usually just by traditional thinking, i.e. observing other people completing the task, then trying to copy the way they did it. The relative social status of the demonstrator to the observer dictates how probable they will consider the solution to be. Few people come up with entirely original ideas about how to do things, and when they do they're quickly copied. An AGI capable of learning from demonstration by humans would need to be able to follow these same rules.
On 29/04/07, Mike Tintner <[EMAIL PROTECTED]> wrote:
Ah.. I didn't get the test quite right. It isn't simply "how many alternative ways can you find of achievng any goal?" You might have pre-specified a vast number of ways of moving from A to B for the system. The test is: " how many NEW (non-specified) alternative ways can you find of achieving any goal?" In other words, my assumption is that, like the human brain, your system's database will know many ways of moving from A to B, that have not yet been used in this connection - ways to move around the world that you have not yet used or specified as ways to move around a home/ office environment, such as "ride a bike", "get pulled on a trolley", "use a scooter" The true test of adaptivity therefore is a test of your system's capacity to find NEW ways of achieving goals from within its database. ------------------------------ This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?&;
----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=fabd7936
