Take a moment and think about what it takes to support your lifestyle. Well, I don’t know anything about you. Take a moment to think about what it takes to support my lifestyle.
Like, I imagine, many of you, I am not a farmer. But, like many of you, I do eat food! I eat quite a variety of food, and none of it is produced within ten miles of me — much of it is probably not produced within 1,000 miles of me. So not only do I need dedicated farmers and ranchers to provide me with food, I also need vehicles, a transportation infrastructure (ie, roads, rails, airports), and then distribution technology (ie, grocery stores and restaurants), because obviously I can’t buy food in wholesale quantities — it would go bad before I ate it. I need shelter, and we also need buildings for those grocery stores (and factories to produce the vehicles that we already need, and houses for the people growing the food, making the vehicles, building the roads, and staffing the grocery stores), so we’ll need a construction industry, and we need raw materials to build those buildings. So that’s a timber and mining industry, and then processing the raw materials, especially from the mines, into things like i-beams and nails. And of course underlying all this, we need energy generation technology to run the vehicles, the mines, the lights in the grocery stores, etc., which in practice means a whole different kind of mining to extract petroleum products (or a different one to extract Uranium-235, or a different one to extract the rare-earth metals necessary for solar cells). And a whole different set of industries to create oil rigs (or whatever). Now, after that, let’s talk about the health care I require…
Seriously, think it through. It’s literally the biggest problem of human civilization. Our greatest crowning achievement is… the civilization itself.
Think about the intelligence necessary to make this thing work. We need people producing innumerable raw materials, and we need other people turning those raw materials into even more innumerable intermediate products, and we need yet further people turning those intermediate products into end products. And you don’t want a situation where at the end of that whole process you have 18 billion iPads and 0 houses, but on the other hand, we do want some iPads! You need a system that trains people to do the innumerable jobs, and efficiently allocates them to the jobs that need doing, and dynamically balances itself to handle a changing world.
There’s a concept in computer science called “NP-hard.” It basically means a problem that requires exponentially more work to solve as the size of the problem grows. A classic NP-hard problem is the Travelling Salesman problem. So basically, imagine you’re a travelling salesperson, and you want to visit several different locations where you’re going to make your pitch, and you want to do it in the most efficient manner possible. Say there are two locations far to the north, and one far to the south. It doesn’t make sense for you to go up north, then all the way back south, then all the way back north again. Instead, you go north, do both of the ones up north, then go south and you don’t have to make the long trip back north again.
So travelling salesman problems are easy enough when you’ve got three locations, or four, or five… but as you increase the number of locations, it doesn’t just get steadily harder, it gets harder faster. Double the number of locations? Four times the complexity. Triple the number of locations? Eight times the complexity. Quadruple the number of locations? Sixteen times the complexity!
Mathematicians much smarter than you or me have studied NP-hard problems (I’m omitting a bunch of mathematical complexity here: they’ve studied NP-complete problems more, but the difference doesn’t matter for our discussion), and while we haven’t completely proven that NP-hard problems are always and forever exponentially complex to solve, we’ve come pretty damn close. Or at least, they’re that hard to solve optimally and rigorously (which basically means: a brute force solution where you try every possible path through the locations).
If you were an actual travelling salesperson, though, you wouldn’t sit down with a pencil and paper and try out every possible route through your destinations. You’d apply some intelligence to the problem and say, “Well, obviously these locations are clustered, so I’ll visit them all at once, and those ones are farther away but in a rough arc, so I’ll start from the top of the arc and walk down it, rather than going back and forth over it.” And you probably wouldn’t come up with a perfect solution, but your solution would probably be good enough. That’s some of the magic of human intelligence: that you can get good enough solutions to problems that are essentially impossible to solve in a straightforward “try every possibility” way.
There’s no formal proof that the creation of all the material goods necessary for our society is NP-hard, but that’s only because we don’t have a mathematical definition of the problem. It is clearly at least NP-hard, and probably in a class of problems that’s actually harder than that.
Now, the good news is that we don’t have to solve the problem of material goods optimally. We don’t need to have no waste ever of any kind. But if our solution is sufficiently bad, civilization regresses to a lower level, and with our present population, that probably means lots of people starve or die of exposure or lack of acceptable healthcare. So the stakes are high, here!
What we need is a super-intelligence. Perhaps not the kind you traditionally see in science fiction, a giant computer with lights, but I think that this is a fairly straightforward proposition: there’s no way that even the smartest person in the world could solve this problem, right? And the amount of just straight up speed necessary to solve the problem by brute force is absurd due to the exponential nature of the problem. So we need something that thinks better than an individual human can, and isn’t just brute forcing the problem. Anything that satisfies those requirements is, I argue, by definition a super intelligence.
The good news is, we’ve got one (or several), since we do, in fact, solve this problem. But I’ll talk more about that in the next post.