I have seen no evidence that the field of meta-learning over AI / problem-solving approaches even has much interest, let alone is showing enough progress that i

I have seen no evidence that the field of meta-learning over AI / problem-solving approaches even has much interest, let alone is showing enough progress that it could foreseeably start self-improving. The "AI" areas where great progress has been made recently are extremely technique- and domain-specific (e.g., image recognition via deep neural networks). Heck, I haven't even seen a lot of progress in areas that seem relatively straightforward & seem like absolutely necessary precursors to strong AI, like automated refactoring of codebases or just straight-up genetic algorithms.