Lots of people think that main problem in AI is a lack of funding. I disagree, Einstein didn't need a pay to work on his theory. GI is the most theoretical problem ever, & such work must be driven by curiosity. Anyway, I offer prizes of up to ~$500K for relevant ideas or stimulating questions & objections. $1600 is paid out so far, mostly for stimulation. This won’t work unless my intro (below) rings a lot of bells for you.
My approach is unsupervised learning, vaguely similar to hierarchical fuzzy clustering. But no method that I know of is close enough. This field is largely experimental (especially ANNs), while I believe in theoretical integrity. So, I had to start from the scratch. I am not interested in combining disparate methods that somehow “work”, any suggestion must fit into strictly incremental-complexity search hierarchy outlined in the intro. Unless you can find a flaw in my reasoning, which would be even more valuable:
I define intelligence as ability to predict & plan, which can only be done by discovering & projecting patterns. This perspective is well established: pattern recognition is a core of any IQ test. A pattern is a set of matching inputs or lower-level patterns. Partial matches & misses are produced by comparison among input patterns, over selectively extended range of search. I see no alternative to these definitions, & expand them to derive operations on the inputs. This deduction is uniquely incremental, hence a singular in “cognitive algorithm“...:
[link][11 comments]