I think that quote could use a bit of nuance itself (albeit, yeah, maybe bringing up Hiroshima like that does set off Godwin's Law alarm). China can develop dat

I think that quote could use a bit of nuance itself (albeit, yeah, maybe bringing up Hiroshima like that does set off Godwin's Law alarm). China can develop datasets on millions (billions?) of people and, yes, given the current state of AI research, that can give them an edge, and I am, too, of the opinion that it's better to not have this edge than to develop it at such an expense in human lives and well-being. So where is the scientific answer to that? Is there any serious additional research effort, and more industrial effort, going into models that are less reliant on training data, for example? What industries are trying to become less dependent on massive data collection, and on massively data-driven models? You can still get heaps of money with a product that can be aptly described as "it does linear regression with sugar on top and it'll tell you great things if you throw more than 100GB of tables at it". Selling an analytics SaaS that sucks up data from every imaginable data source to tell you things about your Google Analytics graphs is still a pretty profitable business. On the other hand, serious researchers in academia, and a lot of talented engineers, are struggling to get funding for topics that do not have the magic words "machine learning", "data-driven", and "AI" in them, and they mostly go ignored because everything can be solved with enough data.