Wakko SicK - Jul 28, 2011:
You are taught something then you use the information you learned in the past from that teaching session to make a prediction that 2+2 was 4 before and so it must still be 4. Hope this helps.
It does help. I guess we have different definitions/views of prediction. I agree that the universe is based on the fact that it is not totally chaotic. An object at rest stays at rest and an object in motion stays in motion. Gravity will hold true from one moment to the next and what you learned in the past should be true in the future. If that is what you mean by prediction, then I understand your viewpoint.
Wakko: We aren’t disagreeing. You are jsut repeating the same thing that has nothing to do with what I originally said. You are arguing what the definition of learning is. That may be an intreating side note but What I said was you can not CREATE. Nothing to do with teaching. ALife teaches itself as shown in the coding of Creatures 3 exodus.
Wakko, are you saying that you believe Alife will be a successful path to creating a “conversational” AI (after all this is Chatbots.org)? I agree that Alife can generate entities that will display intelligent behaviors, but do you think it is a path that will lead to higher cognitive functions?
Wakko: An accident is how we learn most things. This is why teaching ALife would be ineffective in the long term. At every stage of our evolution there was nto the same guy holding you hand. Original thought comes from trying and failing until you succeed. This is why you are not gettign it. Seperate learnign from the equation. You are not gonna teach this thing. You are going to watch it grow. If thats too slow for you don’t try to create any form fo life cause they all have to grow.
The problem with using alife to evolve an Artificial Intelligence that has the cognitive capabilities of a human is that there is no existing environment that enables the equivalent of natural selection. For something to teach itself it must have an environment that successfully generates feedback/response. Without that guidance, alife becomes the Infinite monkey theorem. You might get interesting results, but you probably won’t be able to talk to it.
Merlin said “Even in Creatures, the neural networks are heavily tweaked and tuned during development. This is necessary no matter what kind of model is used for the AI logic.”
Wakko: This is true in the field of chat bots as well.
That was my point.
Wakko: Play the game man. You will be amazed. I find it harder and harder to stop playing and its about 5 years old.
I have played the game. The variations have been around for years.
http://www.ted.com/speakers/will_wright.html
One of the simplest forms of the concept is Conway’s game of life.
If you would like to learn some programming (evolution is then based on your skills) then you might enjoy Robocode.
I have found evolutionary/genetic type of game most applicable to avatar creation/variation (or unintelligent NPCs driven by instinct) and less usable to create an AI that can communicate in a human language (like a chatbot).
http://chrishecker.com/How_To_Animate_a_Character_You’ve_Never_Seen_Before
Now I’ll ask you to play the game too. Go have a conversation with Skynet-AI and then tell me if you believe an untrained/self-trained Alife will be a better solution to creating an entity you can have a conversation with.
Early work in the field of using biology as a model was done decades ago. This is what I was talking about when I said that some people are using biology as a guide for AI development.
http://en.wikipedia.org/wiki/History_of_artificial_life
http://en.wikipedia.org/wiki/Genetic_algorithm
Additional links:
http://en.wikipedia.org/wiki/Development_of_Spore