From: Joshua Fox (joshua@joshuafox.com)
Date: Thu Feb 23 2006 - 13:45:45 MST
About "Levels of Organization in General Intelligence" from the upcoming
book.
At the very beginning of DGI AI implementation, it makes perfect sense
to build modalities on top of toy domains like billiard games. However,
might the next step be re-use of existing modality modules?
Implementations of the modalities as opposed to higher layers are much
more widespread and mature, since the higher layers depend on the lower.
There is a common dream of an ultra-simple implementation in which an
AGI is build out of a very few lines of code. LOGI points out that the
most extreme simplicity is not possible. However, modalities have much
more design complexity than higher layers, and so re-using modality
modules may significantly reduce the total AGI implementation-complexity.
Does this make sense?
Joshua
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT