All this AI risk could be mitigated by not giving them access to any weapons. AI shouldn't have access to military equipment, gases, explosives or anything else that poses large scale human risk.

I mean, if you're fridge becomes a million times smarter than you, what's it going to do? Freeze your food? it doesn't have control over your other devices so it wouldn't be able to do much to harm you.

Good luck confining it to the fridge. Given the horrific security of all our devices, it wouldn't be very difficult for such an AI to leap from the fridge to a nearby internet-connected device and then order itself a few hardware upgrades, including 3D printers, custom-fabricated chips, etc.

> if you're fridge becomes a million times smarter than you, what's it going to do?

Things that you can't even begin to imagine. That's the problem.