Re: A simple AI



On 04/01/2011 2:18 AM, Anatoly Medyntsev wrote:
http://www.numenta.com/htm-overview/htm->algorithms.php
The idea of the numenta's approache is based on some sort of neural
structure. I'm not sure that neural structure give us a hint to
understanding of brain logic. Neuron is morphological basis of brain, but it
doesn't mean logical basis

Regards,
Anatoly
[...]

Kindly set your e-mail/newsreader to "start reply below quote". "WHY?" you are no doubt wondering. Because than one has to skip past your top-posted reply to skim remind one self what you are replying to. Inconvenient and irritating. Don't accept Microsoft's notions of what is good practice.

A few conclusions I've arrived at (for now):

a) A neuron is a network, not an on-off switch. Understanding how it works is a first step (a very small one) towards AI.
b) "Wants" are absolutely necessary for any decision to be made, i.e., for any action to occur. Without "wants", the system will merely compute inferences, but it won't act on them.
c) "Reinforcement learning" has to begin with some built in behaviours
d) "It's conditioning all the way down." (Quoting myself)
e) "Logical" and "intelligent" are not equivalent descriptors. In fact, logic is not even required for intelligence.
f) Effective behaviours are the measure of intelligence. This implies that intelligence is contextual: what's smart in one situation is stupid in another.
g) The brain is not a parallel processor, but a network of simultaneous processor, with massive recursion. The network's topology has at least 4 dimensions.
h) AI is not symbol manipulation.

Wolf K.
.