Monday, April 17, 2017

The Second Tech-Peace-Dividend

The Omnivore has written about what some people call the Smart-Phone-Peace-Dividend. The idea is that the billions of dollars spent on RnD in the Hobbsean war of all-vs-all in pursuit of the next great smart-phone innovation created a swath of modern technology that would never have existed if we'd just gone with minor increments (alright, not never--but you get what The Omnivore is saying).

Basically the tech developed for Smart Phones (cheap touch-screens, better connectivity, better batteries, low-power super-chips, super-good cameras, way-improved storage, etc.) has all resulted in various spin-off technologies.

The conditions for a "world war" of RnD are pretty specific: a new technical frontier, billions of consumer dollars at stake which makes investing billions more in the tech, and multiple players vying for an edge. We got drones out of phones--what comes next?

Self-Driving Cars

The Omnivore sees this happening in the coming self-driving cars product. These sorta-kinda exist today for some of us (Tesla drivers, mostly)--but everyone is working on them. Google, Apple, Uber, and others are pouring billions into this. The key elements of these, computer vision, robotics, and, perhaps most importantly, decision-making capability in 3d space, are all very tough problems that are getting the Smart Phone treatment.

This is very, very likely to produce self-driving cars which drive as well or better than humans sooner rather than later. The Omnivore's question is: "What else will it produce"?

The Robopocalypse

The robopocalypse is the event where robots can replace low-skilled workers at a cost cheaper than the workers and service that is sufficiently good to get rid of them. Replacing truck, taxi, and Uber drivers is an economic MOAB. Replacing the staff of a Wallmart is a thermo-nuke.

The Omnivore had written that when a domestic robot could properly fold a fitted sheet (something The Omnivore does not even attempt) they would have reached the dexterity necessary to replace most challenging domestic chores. The remaining issues, though, mostly around figuring things out, would be left unresolved.

There are generally two ways to approach a hard problem with tech. The first is to make the tech strong enough to actually work the way humans "expect" it to. For example, Google now does a decent job of letting you ask questions of the global infosphere and can figure out what you mean and what to show you by way of results.

The second way is to simplify the problem so tech can address it. An example of this thinking is: "Let's put sensors on all the roads so robotic cars can navigate."

As you may have noticed, this has not happened very often and isn't about to happen.

The real solution is #1 and it's always been #1. Since people have realized this, they are now building cars that can figure out if a person walking near the edge of the sidewalk might step out into the street and then reduce their speed until they can pass them.

Let's look at a different problem of putting away the kid's toys in your house. Maybe you have one of those houses where every toy has its place and you can just put the big legos in the big lego bin and the little legos in the little lego bin and it's all good.

Alternatively you live in the real world and putting the toys away involves things like knowing that the toys go either "in the toy area" or, maybe, "in the kid's room," or, maybe, "in the kid's closet--if it's on it's way out." Possibly: "In the donate-box in the garage if it's really on the way out" or even "in the garbage if it's worn out and broke."

Mom can make all of those decisions. A human cleaner can make almost all of them. A robot? Would you trust it not to throw out your daughter's favorite doll no matter how chewed up it is?

Right. This is a hard problem.

But what if the tech-peace dividend provides robotic vision good enough to sweep up a bunch of toys, pick out the hot-wheels cars from the legos from the random bits, sort them into reasonable bins (the toy cars can all be grouped, as can the legos--the random bits can either go with the toys they came with--or into a random-bin). What if it can figure out when the toy-area is too big and take large items to the kid's rooms (and know the pink ones go to her room and the blue ones go to his--except if they don't).

What if it can identify a broken toy from a whole one?

This sounds absurd from a technical perspective but it isn't. Self-driving cars make very, very different decisions but they don't make fundamentally harder ones. Certainly the stakes for the cleaning robot are lower.

If we can build self-driving cars, we aren't that far from the cleaning robot. If we can build the cleaning robot, we can automate the fuck out of just about everyone who draws minimum wage or anything close to it. Most jobs are much, much easier than cleaning up the kid's toys. Stocking shelves is child's play (yes, The Omnivore meant to)  compared to that.

So if we get those robo-cars in 5 real years (instead of the Fusion-Power 5-years) then how long before the robo-car peace-dividend provides us with domestic robots?

Keep in mind that the robopocalypse is maybe "25 years" away before some experts think we'll have to be providing everyone with a Universal Basic Income since they won't be able to work for the most part. What if that comes  . . . in 10. What if it comes in 6?

Remember that the peace-dividend effect isn't that the market for whatever it is isn't out there--it's that the massive, massive costs in RnD aren't justified to meet it. In the tech-war scenario, the RnD costs pay for themselves because something else that's closely related is worth the costs.

Automating a Walmart (well, let's be real, Amazon gonna get them before robo-workers do--but you know what The Omnivore means) is totally, totally worth it if you can get the RnD for free.

We're getting the RnD for free right now.

The Other Direction: VR

The Omnivore has a VR head-set. It's very, very cool. The Omnivore considers that there's another avenue to the "robopocalypse" which doesn't require Artificial Intelligence: Remote workers. If a worker can drive a robot like they're there, then you have options for pizza delivery guys from Antigua (or any other location where they are cheap to hire--sure, flying drones will deliver pizzas before guys from a Caribbean Island drive pizza-bots in VR, but you know what The Omnivore means).

In this model the cost is for the robot--which has to be close enough to human dexterity to, say, fold a fitted sheet--but after that, having a human drive it can be globalized. Lawn-care? Auto-detailing? Whatever--forget about your local talent--go wherever the (cheap) headsets are.

You also get a video-log of everything that happened so you can audit everything: If that's worth paying extra for, the barrier-to-entry just came down.

When will this happen? Who knows--but VR is gettin' the Smart Phone treatment (with, in fact, Smart Phones as well as other devices) so those pay-offs should be seen in other areas soon (the second wave of high-end VR should start next year and go through a complete upgrade cycle in about 2-3 years, The Omnivore thinks).

The Rate of Change

The point here is that there are accelerants towards some end-games that don't depend on how far the end-game is from the current technology. Just as no one in 2006 predicted the iPhone would give us super-drones, The Omnivore doubts that Google's self-driving car in 2009 is seen to presage whatever the spin-off technologies will be from the billions that are now being spent to put Uber out of business.

The robopocalypse is just the outcome that looms largest to The Omnivore--it's likely that the real innovations in computer vision, robotics, and high-stakes AI decision making will pay off in some entirely unrecognized and highly disruptive technology.

Keep Watching The Skies.

No comments:

Post a Comment