I read through the summary of the report on automation by Brookings and McKinsey. It is worth reviewing because it is a realistic, nuanced view of the dangers and opportunities of automation.
It feels like the conversation is moving closer to what we had before the big advances in neural networks. Those advances, especially Alpha Go Zero, set off a frenzy of speculation that all our jobs were doomed and an AI superintelligence would subjugate us puny-brained humans. I think we are realizing that while the latest advances are amazing, they are still brittle. Alpha Go Zero can’t drive a truck or even tie shoelaces, even if it can whip anyone’s ass at Go.
It feels more likely that ten years from now we will have really powerful pattern matching machines that can do things like drive trucks and tie shoelaces. Yet those machines will probably suck at playing Go.