| > Any unskilled job... That's a moving target with software in the economic picture now. We know about the "That isn't AI" effect. Each incremental advance coming out of AI research or just software in general is countered with a problem domain the advance doesn't address, and the advance is dismissed as not really ground-breaking. This effect is commonly seen in action with traditional games (checkers, chess, Go, etc.), but I see it with all sorts of automation. I posit there is a converse rule: every incremental gain (which admittedly are infinitesimal steps towards AGI/strong AI) has the potential to move the goalposts of what constitutes an "unskilled job". We often fail to recognize this because the incrementalism's effect upon jobs is not nearly as dramatic as what we today imagine the transition from horses to cars to be (when in fact it was quite incremental back then as well). There aren't as many checkout cashiers as before due to software- and infrastructure-enhanced advances (bar codes and their scanners, supply chain management, etc.*), but because checkout cashiers aren't completely gone yet we don't viscerally perceive the economic effects. What used to be a skilled job is now classified as unskilled however, so the economic impact is huge. I contend we don't need AGI for massive automation-originated social disruption. If someone produces software that roughly approximates the reasoning abilities and kinesthesia of a 7-11th grader, then that is the majority of people in the developed world, not to speak of the developing world. That developmental range is roughly equivalent to what the mainstream media content targets in the developed world. We'll see continuous, smaller disruptions far before that point, though it is an open research question if we will get there in the near future. For example, we still don't have a general software solution to a task as "simple" as folding clothes, sheets and towels grabbed at random from a dryer. |