AI isn’t a being. It’s a road.
I have spoken of the confusion about robots and AI before. Because people believe that we are on the cusp of making new beings in the universe, and have compared AI to aliens or a new forms of life, terrible ideas are propogated – like the handwaved Singularity or the even more handwaved “everyone will just train for new jobs” or the amazingly bad idea idea of robot bosses. And I don’t even want to mention this charleton who wants to create an AI God. That idea alone is going to cause immense pain to many vulnerable people.
You see, we have the metaphor all wrong here. That’s the heart of the confusion, thinking that machine intelligence equates to organic being existence.
We are thinking that AI is like a human. But it’s more like a conduit of information. A road.
(And yes, “AI is a Road” is a metaphor, I know it’s not actually a road, my argument here it that there are many similarities to AI and roads, the first of which is that, like a road, it’s a piece of “invisible technology” that nobody thinks of as technology anymore)
Neural networks, as in our brains, are just reinforced pathways in a collection of interconnected points. In fact, this is what information itself is. Those links between nodes become more or less weighted as the paths are exercised. The connections find signal in the noise. A dirt path turns into a road, a road turns into a highway, a Starbucks is built, a hipster buys coffee and blogs about cosplay.
No matter how well you build a road, it’s still a road. The road itself should be based on where people need to go. We don’t build a world designed for roads, we build a road where a path exists already. In the same way, AI research should not be looking for jobs to replace, but rather for ways to improve our work and home lives. Ways to make us even more human.
We must not let anyone create needs for AI. I propose that doing so is a crime against humanity. In my First Law I state:
All research into and attempts to tie AI behavior into systems which simulate rewards and punishments shall be deemed a serious safety hazzard. This would be an offense on the felony level, as it could result in widespread, catastrophic damage.
If AI is a road, then building an AI with needs is like building an unsafe bridge that threatens travelers along the road.
And who determines if a bridge is unsafe?
It is time for engineers to rise and take control of this process. The politicians, with a few folks like Andrew Yang being exceptions, are hobbled and distracted with the games of power and dominance. Computer engineers must work with social engineers – lawyers – to build systems of restraints for public AI.
I think orgs like the Partnership on AI are moving us in the right direction on this discussion.
We must make laws for our new roads. Or toes, and worse, will be squashed.
Painting by Simon Stålenhag