Sling to Net AI.
This is the type of AI where your trying to sling real world data or generated data into a tensor the net in other words. From a dynamic sling and inter-connected tensors you can auto generate evolutionary GANN challenges(the net could bounce and wobble) relative to the data being sung and the learning challenge roster.
More general AI may come from this closer to real world aware vehicles and bots.
Hook and Sling AI.
This type of AI is where you constantly keep refining the hook. The bit you have to calculate with variable efficiency and refining the sling the bit you can add to the output through inferring to make it all look better.
Video's and Game's should improve a lot thanks the reductions in calculation possible.