Artificial General Intelligence Analogous to Flight
I think it might be worthwhile to consider the development of flight as an analogy to how to develop Artificial General Intelligence. Consider:
http://www.loc.gov/exhibits/treasures/wb-timeline.html
Timeline of Flight: The Dream of Flight (A Library of Congress Special Presentation Commemorating the Centennial of Flight)
What I see coming through is toys being used to develop flight like the balloon being made larger to form the hot air balloon or the kite being made bigger to hold humans.
It’s incremental:
balloon->hot air balloon -> steam powered balloon
kite->glider->powered glider
They are taking something that exist that does partially what is needed and expanding its properties [ie make it bigger]. Next, they improve the performance, make it faster.
I think we have passed the point where we make it bigger. For instance, neural nets can do some task we want, but it requires a lot data and it is slow to train. The same thing with Genetic Programming, it can develop new inventions, but requires a huge population which in turn requires enormous computing power. Then there is brain simulation, we can do that for a mice brain: http://news.bbc.co.uk/2/hi/technology/6600965.stm , but again it is too slow.
I think one of these 3 approaches will definitely take the cake of finding Artificial General Intelligence. We are at the stage where they need power to make them very useful.
So, I think we need to exam ways to improve performance of these algorithms:
1. Hardware — we could develop faster hardware.
2. Improve our algorithms
1. For us ie dont know, can we design new chips. Where do we get the money and facilities?
The path of least resistance for us is through improvements in our algorithms.
How can we make genetic programming faster?
How can make neural net require less data?
How can make training neural nets faster?
How have we made other algorithms faster or require less data?