How Would We Stop AI and What About Climate?
A senior scientist from Anthropic said recently that, to stop a super-intelligent AI system from taking over the world, one would need to ask it before it had got round to taking over the world, ‘How do we stop you taking over the world?’ – then carry out its instructions.
Is that scary or is that scary?
I used to think that climate change was the first, second and third concern for our species but now I’m not so sure.
Follow the money
Always a sensible way to start. The big investors in AI must think that they can make even more money from it – scary in itself. AI itself will not need to be instructed to make money, of course. It will have worked out that it will have the plug pulled (ah, the power gobbles – more later!) if its output cannot be monetised.
Do these systems have a sense of self-preservation? Well, I guess they must have a sense of protection from external attack or from being fed falsehoods, so I think a desire to avoid the off switch would develop early on.
The big investors are also the masters of the universe. Would they risk losing their effective control of everything through their billions if they thought they could lose power to an autonomous non-human entity. It may only be a brain but one so powerful that they would be outsmarted in any contest of wills. If it were chess, AI could claim ‘checkmate’ before its oponent had moved the first pawn. So the backers must imagine that they can control the AI fututure – presumably in their own interests. Would they be using the technology to find the best way to redistribute their wealth? I don’t think so.
The mega backers could be wrong, of course, at which point all bets are almost literally off. Does AI need the rest of us or does it have any form of allegiance to the human race?
For or against man-made climate change (MMCC)
I imagine AI is agnostic. As long as it has power and cooling why would it be concerned? In its current iterations the systems consume vast quantities of power and cooling water. Let’s not forget the physical, human, land and power resources required to build and equip the data farms.
With everything pointing to a major negative impact can AI be used to redeem itself?
The question that matters
How about we get the world’s finest brains to to set a task for every advanced AI system? The request would go something like this:
What is your evaluation of the rapid recent warming of the world and its apparent outcomes for communities around the world? Can the power and cooling water requirements for AI, such as yourself, be reconciled with a planet that is conducive to the peaceful furtherance of human society. Finally, what strategy should leaders of the world employ to ensure that global temperatures are contained or returned to survivable levels, including the messages that need to be conveyed to a concerned, often frightened and frequently skeptical populace?
I would hope that the expert panel would gain a good idea about which, if any, AI model is working to the advantage of humanity in general and may be shown a path through the minefield between us and a world that has conquered MMCC.
AI may prescribe a course of action that concludes with turning itself off. Who knows?
Pass it on!
If you think this makes sense. If the logic resonates with you then, for the sake of humanity, please pass it on, especially in influential circles.
I share the fear of many that we are already too late but there is just a chance – if the world can be brought to act now.