Greetings Telmo,
I've responded to your comments below.
Are you working on an ontology based AGI approach?
Stan
On 04/28/2014 02:30 PM, Telmo Menezes via AGI wrote:
Hi Stanley,
On Mon, Apr 28, 2014 at 9:23 PM, Stanley Nilsen via AGI
<[email protected] <mailto:[email protected]>> wrote:
Hi PM,
A few thoughts -
One might try to come up with methods to generalize situations -
put in categories and sub categories and sub sub categories...
This sounds logical, but also terribly tedious.
My alternative is to look at the world as sets of triggers. A
trigger initiates an action - maybe to assert a new fact. The new
fact might then trigger something else...
Ok, but I don't see how this removes the need for an ontology.
As I understand it, there are several efforts to create massive
ontology. And, we can all see the "value" of it. The struggle is in
finding the mechanisms that can cash in on that value - the magic sauce?
I focus on how to become more intelligent when you start at next to
nothing. What's the bootstrap look like? At what point does a computer
begin to build it's intelligence? And, what do the construction
elements resemble?
It could be implicit or explicit, but you still have to be able to
make more and more distinctions between triggers or actions. I tell
the AI to book me a trip to Cambridge. What Cambridge, UK or USA? And
then, to book the ticket I have to know that Cambridge is a town, and
that I already know something about how to book travels into towns,
and so on.
Software "assistants" are pretty popular now. I understand Microsoft is
planning one to compete with Siri. Maybe this is the way to the
future. Start out assisting and one day take over :)
You need some way to generalise, and this leads to some hierarchy of
types. I bet our brain encodes a huge one. But how does it encode it?
What is triggered depends on what our "understanding" makes of
triggers. Pretty much a Rube Goldberg contraption, but gets
interesting quickly. Understanding isn't that vague, it's whatever
can be coded into rules.
So you would say that a thermostat understands temperature?
No, I would say that whatever is reading and setting the thermostat
needs to understand the effect they want to achieve. The "user" chooses
the thermostat based on understanding of outcomes that are expected.
The thermostat is simply a "see" mechanism - it triggers something
else. If you wrote a rule to act like a thermostat, I would say that
the rule understands an aspect of a thermostat - e.g. numbers change
over time and there is a trigger point. I don't think the rule needs to
know about atomic vibrations, or the cost of a barrel of oil.
I'm not downplaying ontology, it will be useful. I just don't put it as
first priority in building an AGI.
Stan
-------------------------------------------
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
