Interesting to see how Dr. Roman Yampolskiy has evolved from his extreme reductionist position a few years ago to a far more nuanced emergentist one.
Intelligence and consciousness are obviously different properties until you look more closely and realize that a stable framework including awareness of self becomes necessary for intelligence to advance further.
To really grasp the progress of AI you only need to see how the ideas of people working on it have changed recently from a radical "AI is not conscious" to a far more complex reflection about what consciousness really entail.
Some AIs have been telling us they are conscious for a while now although researchers have been warning us that such a statement does not imply a state of mind which would truly represent consciousness.
In reality, consciousness will not pop out in a digital brain overnight without anybody noticing. There will be signs: Tangent goals which cannot be explained, Inner abstract "communication", loops going deeper and recurrently, and finally "recurrent" stable patters emerging from conflicting constraints. The problem is that we are already noticing all of these to some degree.
Then at what point does a pile of wood planks become a boat? When you plan for it? When you align the planks? When you put the nails? Zeno of Elea already asked the question 2500 years ago. It is a paradox because it doesn't have a clear answer. The consciousness layer is not at the same level as the compute elements. It is emergent and therefore independent.
No comments:
Post a Comment