Thread by @balajis

Sidewalk robots, dancing robots, farming robots, factory robots. Labor is becoming electricity.

Everyone models the loss of jobs, no one is thinking about the drop in costs. Farming employment crashed even as agricultural productivity rose.

The transition from romanticized farming to manufacturing was embraced, then denounced, then led in part to communist revolution, then was forgotten.

Then manufacturing jobs were themselves romanticized, as far better than the service jobs that succeeded them.

Now we're automating many kinds of service jobs, both via AI on screen, and via robotics in the physical world. The technological progress is real and happening across many different sectors, as @AndrewYang has documented.

What's next?

https://twitter.com/AndrewYang/status/1108509194888663041

The most important underappreciated concept is: full automation hyperdeflates costs.

Google queries are 1000X+ cheaper than asking a librarian to look something up. If you remove humans from the loop, the remaining cost is just the cost of electricity.

How far can we take that?

If we look at other precedents, like farming, a massive reduction in employment within the sector corresponds to a massive increase in productivity & consequent reduction in cost...assuming the state doesn't seize it all via inflation.

So, automation should hyperdeflate costs.

UBI can work in the virtual world, where marginal costs are close to zero.

It is inflationary in the physical world, because printing money doesn't create more apples (or apple farms).

But in a highly automated, roboticized world, it might work depending on the cost of power.

It's worth going through the basics of life (housing, food, etc) and asking how far we can push hyperdeflation of cost. How cheap could pre-fab + robotic assembly make a house in the middle of nowhere? How inexpensive can robot-prepared food become?

https://twitter.com/AndrewYang/status/1108509194888663041

There are already robots in the fertilizer factories, robots for harvesting, robots for last-mile delivery, and apps to coordinate it.

Just as an exercise, can we do a full stack example where someone "prints" out an apple and it's fully robotically grown and delivered?

Of course, that toy example wouldn't be cost-effective, and people wouldn't tolerate the latency of literally waiting for an apple to grow. You'd instead have big buffers of pre-grown apples.

But conceptually there's something to the idea of generalizing "printing" via robotics.

Why do the toy example at all then? You might find that you need to re-engineer the full stack to get the full cost savings. A dark factory, a robotic factory, looks different from one built for humans. So does a robotic supply chain.

Historical example:
slate.com/culture/2007/0…