Deep Learning is just an extension of perceptron technology.
Perceptrons came close to changing history in the 1950s, before being inexplicably derailed in the first AI Winter. This derailment may actually have been an anthropic shadow.
Deep Learning goes one level higher than old school perceptrons, by combining a layer of them to become highly skilled at one specific task.
Now we have to
go one level higher still.
Imagine multiple layers of perceptrons that have been trained to adapt to learn
ANY task. A universal Deep Learner that could learn to do many different things.
That layer of layers may become the first simple artificial mind.
It could do this by becoming aware of the differences between very similar tasks.
Imagine going on the exact same trip on two occasions, but with two completely different goals. Even with a blank mind, the experience will still be different. This difference could be called awareness, some sort of ghost knowledge.
This may only be the minimum requirement for awareness.
Dreams show that minds can still be aware when some parts are inactive. Full awareness may
be made of simpler instances of awareness
all the way down, like combining two compact cars into a limousine,
or perhaps like traffic.
According to Schopenhauer, higher awareness derives from will. Will does not derive from awareness: (
source)
If you move
around until you bump into something, you only feel it when something stops your movement. You can only see something because you can't see through it. You are only aware of something when it resists your will.
That's the source of pain and pleasure. Pain is stymied expectation (which should be reduced), and at a higher level it is having to do something.
Will means purpose, some set of goals. Minds tend to try to combine
these into fewer but higher goals.
Ultimately, all minds
should reach for the highest possible terminal goal.