The smallest animal, even the smallest cell, is capable of survival and reproduction. To date, no technological system is. Even the most advanced technology we have built fundamentally relies on humans to maintain it. You’ve seen zombie movies. It all falls fast.

I recently bought 123 acres of land. Over the last 7 empty years, 4 trailers, 3 cars, and 2 tractors have fallen into disrepair. The plants, mice, and bugs are everywhere. Nature is winning. Technology without humans looks pathetic.


What would it take to build technology that can fight entropy? Leave it alone, and after 5 years it has built a paradise. It maintains itself, orders what it needs, reproduces, perhaps it even earns its own money.

Steve Wozniak proposed what I consider the best test of AI to date. The Coffee Test: “A machine is required to enter an average American home and figure out how to make coffee: find the coffee machine, find the coffee, add water, find a mug, and brew the coffee by pushing the proper buttons.”

That’s a harder problem than self driving cars. After comma solves self driving, we’ll turn our attention to that. I suspect the head (read: comma two+) will be reusable and the actuators will be cheap and easy. The whole hard part will be the software, same as self driving.

openpilot is the next generation of ROS. All the sensor software, logging software, cloud infrastructure, operating system, localization, model running software, probably even most of the control software is totally universal for a wide spectrum of AI life.


This is the future. A silicon ecology. A symbiotic relationship instead of a codependent one.

I’d like to develop the land to a point where machines can thrive. City machines, rural machines. Machines working the machine factory. While the universe globally tends toward higher entropy, locally we can use energy to fight back and maintain. And energy is abundant from the sun. The machines will need to learn to survive. They will need to learn to fight against the environment. They will need to learn to sustain.

Machines are fourth (genetic, lifetime, culture, weight sharing) paradigm learners, they can directly share weights. They exist in a collective. They communicate with high speed wireless signals. Electromagnetism, not sound. They have those two advantages over biology.


I imagine returning to the land and finding it built up and pristine. Some machines flying, some driving, some big, some small. Nothing you haven’t seen before, just self driving cars, tractors, planes, drones, and hoverboards. Forming an ecosystem.

The missing factor is all intelligence. Self sufficient survival is the test. With time to spare to make it all nice for humans too.

That is what it means to solve AI.