Widespread poisoning attacks making it harder to compile good training datasets. 

Availability of data to train on is a bottleneck

If you have suggestions for a reason other than those already available as an option, please comment and I will add it. I am not allowing free-response to mitigate spam of low-quality answers.

Resolution will be to either YES/NO for each option that qualifies if it is obvious, else resolves to PROB (default : 50%) for an ambiguous "sort of but not fully" situation. 

Multiple slowdowns can happen. For example, if supply chain issues cause it but we recover from those and then start running into hardware bottlenecks, both of those options will be resolved to YES.