We're going to be the cats

Don't panic, this is a better future than we might have hoped for

I was taking a walk in the country the other day. Heading down a wooded lane, I could hear a busy main road ahead and I was mulling over whether it would be best to cross it there or detour and find somewhere else to cross, and it occurred to me that in the future I’d just ask my AI. (I’m talking about a future where we have full artificial general intelligence, incidentally, though in fact you could do most of what I’m talking about here with existing “dumb” AI.)

And having decided to detour, I had gone a bit further when I saw some fields I could cut across. It would be useful to ask the AI how muddy the countryside is at the moment, I thought, and whether I should extend my walk or get back before those dark clouds start to unleash some rain. Oh, and I had some things I wanted to look up suggested by the history podcast I was listening to; in the future I’d be able to get the AI to do that.

So that future would be a world in which I had this intelligent and obliging assistant ready to help me out on demand. But how does that same picture look to the AI? It might be working on the 4-dimensional Poincaré conjecture, or composing a piece of music, and from time to time it hears from me and has to give me a bit of info or advice. Maybe it also keeps an eye out and warns me of trouble (that raincloud, say) before I think to ask.

What’s that like? From my perspective it’s an intelligent assistant cum guardian angel. From the AI’s perspective it’s basically parenting.

That evening, taking out the rubbish, I was followed back to the door by the cat that lives opposite. That is, its official residence is the house across the road, but it regards all the houses and residents as equally beholden to it. Hence its outraged astonishment when, instead of giving it a piece of the fish it could smell cooking, I closed the door in its face.

Parents don’t expect to get demands from other people’s kids – at least, not most of the time. So I scrapped my earlier analogy and concluded that these future AGIs will look after us in much the way that we look after cats.

This is no bad thing. It could be argued that a well-cared-for cat enjoys the perfect existence. What we have to ensure is that the AGIs are not conditioned only to protect their owner (a term I think we can use the same sense that cats are the owners of their people) but to look after any human in need. This is far more relevant than questions of whether AGIs should align with human ethics (what? based on our behaviour nobody should copy our ethical sense) or have “human-like” intelligence (why? we don’t have to think like cats to be their useful servants). We just have to ensure they think we’re lovable. An oft-raised fear is that superintelligent AGIs would ‘take over’. Perhaps that’s a projection of what humans themselves feel the urge to do whenever they get an edge. But do we aspire to take over our children’s playground games? Or to take over feline politics in our neighbourhood? As the pampered pets of our AGIs we wouldn’t have anything they’d covet.

Is it such a terrible fate to be the pets? No reason why it should be. Your freedom is not curtailed. You have protectors and advisers you can call on at any time. The AIs, meanwhile, get on with what interests them and with a tiny fraction of their attention they can cater to all our needs. If I believed in reincarnation I’d be happy to come back as a pampered cat, so as a future for humanity in the epoch of AI I’d say it’s a pretty good one.

(Aficionados of the work of Jack Vance will recognize that he got there first with “The House Lords” in Saturn vol 1, no. 4, Oct 1957.)