I just don't buy it. I understand I'm in a minority, so let me indulge myself and explain, for a moment, why I think AI / Robots are nothing to worry about. My claim is: Robots will not replace humans in all jobs. Further, it's more than because of Human preference, I think that Robots will lack cognitive abilities that we have. My expectation is that in each area where they are deficient then they may well become a…
An "80 IQ" manager will fail at the list of six tasks, and a "120 IQ" manager will be a success, by most peoples analysis.
Surely, an intelligent but inadequate AI would be mentored like an intelligent but inadequate human manager. We have millennia of experience proving humans can't do that successfully. Therefore if we invent an inadequate AI, regardless of being close to human performance, we know we can't improve performance to make them useful. We play games with re-rolling the dice, but the clock time and human labor time make that too expensive (although it works for mere numerological metric number optimization problems). We play games with pretending that the history of flipping honest coins implies something about the honest coin although it implies more about self delusion and statistics. We play political games where the leader is useless but at least the leader is our guy, our kind of people.
Sure, for the sake of argument, I'll take it on faith we can build a mentally disabled AI manager. But we know based on centuries of experience we have no idea how to "fix" a mentally disabled manager into something successful. This is not necessarily a complete black pill. A developmentally disabled AI might, if its merely disabled and not evil, be handy to provide minimal supervision to small children or animals in a safe controlled environment. Its just not going to be very useful.
Maybe rephrased, that's nice that some believe we can run a marathon, and maybe we can get part way, but if we absolutely know based on substantial experience we can't upgrade an athlete to run the critical last 100 feet to break the tape and finish, then why bother? Or if we have to build a bridge 1010 feet long if in separate experiments we know we can't extend a bridge past 1000 feet long then it doesn't matter if we have faith or belief that we can build it 10 feet or 950 feet or 999.9 feet because we know once we hit 1000 feet we have no idea how to get it to 1010 feet based on centuries of trying really hard using off the shelf cheap self reproducing biological bridges of 1000 feet length.
My greatest fear of AI is it will turn political and we'll no longer be able to talk about the science and technology of AI; it'll just be another dog whistle signal for one side or another, like many other topics "discussed" here on HN. Kool kids will have the right blind faith, the enemy will have the wrong blind faith.