>>1058900 (OP)
Honestly just adding more philosophers and neuroscientists and cognitive scientists into AI research would be way more practical. That's the actual issue at hand, providing inspiration to where programmers and mathematicians should focus on algorithm development to capture more elements of cognition. The only diversity that's relevant is diversity of ideas spanning from a range of fields relating to intelligence itself.
On another note, most of the recognition problems fix themselves if an AI is given the capacity to self-regulate visual perception by context for learning. That means the ability to focus attention on certain types of abstract data, while filtering out all the rest for learning/executing sub tasks contextually. Then with that data, it can be used to meta-learn how to compare and relate abstract groups of sub-tasks into concepts. The result is that it'll eventually learn to recognize all faces by figuring out which sub-tasks are useful for facial recognition and how common abstract concepts are present in all human faces.
You'd think it would be more useful to figure out how an AI can deal with limited data, and reason for itself to recognize similar things instead. That's what really pisses me off about the article, it offers absolutely nothing useful to address their own concerns and poses a serious threat to AI research if taken seriously.