Artificial General Intelligence (AGI) is the next level.
It is my proposal that AGI will look very little like the current AI paradigms. For AGI to be successful, it has to look much more like an animal nervous system (i.e. Physics) and not the mathematical model that is Deep Learning et al.
Embedded hardware AI algorithms
Advanced software solutions
Better NN (neural network) architectures
Better and robust interoperability tools
Advanced bias checking in AI
Better and understandable XAI (explainable AI) tools
The newer discipline of neuronal network science is revealing a lot of information that is giving us insights into how to create node and edge systems capable of deriving animal like intelligence. Using the neuroscience, one can create technology that mimics neuronal systems. Using spiking neural networks concepts with the added technology to strengthen/weaken connections as well as add/remove connectivity between nodes, allows for artificial plasticity.
We can, and currently are, creating artificial connectors with interesting results. Living, inorganic beings are being created and developed as I am writing this answer.
Current AI tech (e.g Deep Learning et al) is very useful but has reached its technical limits as is. Yes, we can add more computing power to make AI more robust, we can play with multiple DLs and mix and match them to get even more interesting results but why bother if we have the blueprint already available = animal nervous systems? Its as though a rather large number of mathematicians (aka computer scientists) are looking for a miracle breakthrough where none exists because the approach is wrong.
I could say so much more but it would be a book so let me leave you with the thought to look at motor output as the primary source of our intelligence and not sensory input. Current AI is all about sensory input and processing what that input means. Realize we wouldn’t need a brain if we didn’t have to move.
.jpg)
No comments:
Post a Comment