The risk isn't skynet. It's insufficient review and oversight of algorithmic decision making.
Google "robodebt australa" for weaponised governmental hounding of poor people from bad software and multiply by any buzzword deployment of an LLM or GAN.
AI is dangerous because people believe it as well as believing in it.
AGI isn't happening. It's hype machine lunacy and what-if headline chasing. What I read, Hinton is quite careful about how he couches his risk concerns.
If AGI does come then humans can basically call it a day. Which government would not put a AGI in charge? And that will cause an avalanche with other countries following sort. After that humans are no longer in charge of their destiny.
If AGI does not occur, then we should accept AI as a neat tool that makes mistakes. Which means it's basically a bubble (see block chain technology) as with all new technology that comes along and that we initially don't - fully - understand. Eventually we organise society around the new tool and life goes on.
You lead with an IF which is what most pop Sci ends with. As IF goes, it's like fusion.
You end with we adapt: the aim of encouraging regulation is to avoid foreseen pitfalls. The problem is getting mindshare so people limp back to skynet because it's getting eyeballs.
Really, the key points are regulation, legislation and taxation. If "AI" no matter what it is becomes vested with corporate person hood and can file patents and own real property, the societal pains won't be about AGI, they'll be about how rich persons leverage it to their advantage.
If "AI" no matter what it is undermines integrity in decision making we wind up with stupid decisions.
Then predict the future and tell me what will happen tomorrow. If you can't do that, then it's an if.
Legislation and regulation is not the solution. It makes the assumption that those in charge have our best interests at heart (spoiler alert: they don't.)
I put up two options for the future, you choose, I don't care about your future nor your choice. If you believe legislation is the answer, go for it. I don't.
Google "robodebt australa" for weaponised governmental hounding of poor people from bad software and multiply by any buzzword deployment of an LLM or GAN.
AI is dangerous because people believe it as well as believing in it.
AGI isn't happening. It's hype machine lunacy and what-if headline chasing. What I read, Hinton is quite careful about how he couches his risk concerns.