Semi-Retirement

I was sitting at lazing on the veranda here in Goa slowly getting baked in the noonday sun and contemplating just how fortunate I feel. Life has been good to me. It’s hard walking these streets and hearing the women beg me to just buy something for 20Rs ($0.40) so they can buy food for their family.

Fortunately, I’m pretty sure I’ve worked out a solution. I can feel OK about occasionally doing nothing more than lazing about if we just end poverty.

I think I can explain it simply enough that most people will understand. It’s not a complex idea, but it does require some understanding of Computer Science…

If you were dressed in the morning by reinforcement learning clustering algorithm you would arrive nude in a room with a deaf mute butler and a huge pile of garbage. He wanders over to the pile and returns with a banana peel. You scowl and throw it off to the side. He returns successively with a balloon, puppy and pack of gum which you throw away with derision.

Butler

Then he shows up with an army boot. Not your style, but you feel he’s really trying, so you just scowl a bit and toss the boot off to the side. He heads off and returns with a stylish sneaker you like, but it’s bright green. You frown and set it on the floor. He returns with the same shoe in a respectable color and you put it on.

This process repeats, but the next time instead of starting with something random like a marmoset, he brings you pants. They forgot to tell him that he was a dressing butler, so until you picked out the shoes, he didn’t presume anything from your nakedness.

The first few days he brings you clothes, his taste is a bit eclectic, but little by little he starts to learn your tastes. If you don’t particularly feel like you don’t want to pick an outfit, he’ll put together one you like pretty well. And when you feel like you’re in a rut and want a new style, he always brings you interesting things.

That’s all reinforcement learners do, they are constantly performing little experiments to test the statistical patterns it thinks it has found.

All we really need to do to end poverty is to create a virtual world that contains the salient characteristics of our physical world and set a reinforcement algorithm to work on optimizing self-reported life-satisfaction.

When I feel like I’m tired of being semi-retired as an inventor, lets make a computer system that can give me a hundred possible jobs across the world I might like. Let it examine every city, every office dynamic, every match of my skills and the requirements of every job. When someone is hiring, lets make it so that they see the top hundred candidates on the planet.

We could have a world without borders where we declare that the most fundamental human right is that of self-determination.

The trick is the programs need data. This will be a data collection project on an unheard of scale, but it is doable. That’s the point of Project Mímis, to enable to reliable and open collection of this data.

With it we can get four datasets: what people think about each other, how individuals think about the world, what requirements jobs have, and, finally, how satisfied everyone is with the things in their life so that the programs have something to maximize.

I think that some people fear losing their freedom to the machines, but how about we let our children deal with the problem of whether or not to ask for the list of the dozen most compatible singles in a 20 mile radius of their house. All that the individual is trying to do all the time is be honest with themselves and the system so it can properly maximize the values.

Leave a Reply

Your email address will not be published. Required fields are marked *