I look forward to part 2 and suggest a part 3: predictions from 10 or 20 years ago that haven’t come about – and why.

The question is do we have to build on old assumptions and simulate down to the synapses level? We know that the problem is a little more complex than that anyway. How is BAC[1] (Back propagating Activated Calcium) reflected in a memristor? Or all of this [2,3] ? Do we have to do it on that level or are the math principals reflected in higher level abstractions ? Priming a neuron can be found in high level decision making[4] for example. Which is equally important, as a math principal, in voice recognition and understanding, vision and visual illusions (uups) and complex problem solving (teaching a machine math).

In other words all this HW is pretty useless if the principals of a smart system are not reflected in the SW.
Like trying to build a touch system without changing the underlying SW assumptions. Which took over a decade to resolve. Or do I need a 3d transistor to make an dumb system faster, or do I use better sensor integration to give the system a touch of self and make it smart? I for one don’t want to be bombarded with useless crap, I want a machine to reflect on me and provide information when and what I need. How can a machine reflect on me without “self”? Do I need a data center for that or is it a rather simple principal reflected in [4]?

Just some thoughts from the SW side.

1. http://www.jneurosci.org/content/26/41/10420.full
2. http://scienceblogs.com/pharyngula/2010/08/ray_kurzweil_does_not_understa.php
3. http://spectrum.ieee.org/tech-talk/semiconductors/devices/blue-brain-project-leader-angry-about-cat-brain
4. http://www.wired.com/wiredscience/2011/06/brain-risk-probability/