Subsistence farming? I think anywhere where it is simply not worth putting AI will be the last to go. The easiest industries to replace will be first (factory w

Subsistence farming? I think anywhere where it is simply not worth putting AI will be the last to go. The easiest industries to replace will be first (factory workers, truck drivers), I think the high value jobs will be next (simply because they are high value, there will be more effort in replacing them) e.g. managers, ceos, engineers. (these will be a little ways off though) There were a few papers published this year using reinforcement learning to optimise architectures for CNNs and recurrent neural nets, I think this sort of thing will only get bigger. If you think about it, researchers just do a bit of guided random search, something that reinforcement learning can do pretty well. So the 'research' job title may describe applying these AIs instead of actually doing low level research.