As long as we’ve been making computers, we’ve been trying to make them beat us at chess. That sounds like an odd thing to do with a computer out of all possible things that can or could be done with one. Until you spend a little time figuring out how one makes a computer that can beat a human at chess.
And then you get it. And then you’re scared to death.
We had two options to consider when we made the first chess computers nearly 70 years ago. The first one was to have the computer try to select a small amount of optimal moves, say four, and evaluate the outcomes of those moves and select one. The computer accomplished that by evaluating a decreasing amount of follow on scenarios, say three for each of the four initial selections, and then two for each of those twelve selections, and then one for the next 24 outcomes. A computer fifty years ago could evaluate 24 positions reasonably quickly, in about as much time as you have to make a move in a chess game. So that’s what they did.
The second option was less elegant: evaluate all moves on the board as far out as the computer could in the allotted time. And then choose one. It was a method chess world champion and eventual opponent to IBM’s Deep Blue Garry Kasparov referred to as “brute force” in his fascinating book Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins.
In the 50’s, the option mentioned first above was more effective. Computer processing speed was so slow that brute force calculations weren’t nearly fast or thorough enough to add much value. But as processing speed increased, things changed. The first option had a problem. The computer wasn’t nearly as effective at narrowing down the board to those initial four choices as we were. They could calculate the outcomes further in advance than most humans, but they were worse at choosing where to start. Human experience and a broader understanding of the game made us better than the machine at that. So as processing speed advanced, the brute force of the machine took over. Eventually it was too much for us to handle.
When Kasparov, then the reigning chess world champion, lost to Deep Blue, he lost to a machine what was able to evaluate 200 million positions per second.
That was twenty years ago.
Today, computers have the ability to calculate all possible moves on a chess board if it were reduced to only seven pieces. Meaning that by the time you get to seven pieces left on a board, the computer is all knowing from that point on and cannot be wrong. Ever. The work is ongoing to increase the computer mastery to more than seven pieces. Eventually it will get to all 32. And we’ll have “solved chess”. Not the way where a computer is better than the best human. That happened two decades ago. We’ll solve chess in a way that the computer, in as much as chess is concerned, is god. It knows all and sees all before it happens. Instantly and simultaneously knowing all futures is the same thing as knowing the future after all.
If you’re wondering why we’re not further along on solving the whole board, it might help to scope the size of the problem. There are as many potential chess moves in any game as there are atoms in our solar system. It’s not a small problem. But we’ll get there easily within my lifetime.
Which means one thing. And it may seem like a leap. But if you know how this stuff works, it’s not. We’re a generation away from computers doing almost every job that we do today, better than we do. If you’re a skeptic, and you’re not quite willing to go full Skynet or The Jetsons, then maybe you can make this much smaller leap then. From today on, as technology advances, computers will either be able to do more parts of existing jobs, or more jobs all together. Which means, in a world where free markets dictate investment, it’s certain that there will be less of us. And more of the machine.
Almost no one is immune.
Think about the size of a police force with no requirement for traffic safety enforcement. Think about the aviation world where the pilot is a backup. You may always need one. But will you need two? When a doctor can see three times the patients in a day than they can now because intake, initial diagnosis and examination are done before a patient gets to the office, you will simply need less doctors. When the task is to review, approve or override a prepared diagnosis, that job looks quite a bit different than it does now.
Those are the jobs considered highly durable because of skill and human judgement requirements. If you do anything that resembles a repeatable task today, and you plan on doing it for the next thirty years, you might want to think about another plan. The somewhat frightening truth is that, from a technology perspective, any job that involves someone performing a task or making a decision that’s been made in similar circumstances before, can be done at near parity or better than humans today with existing technology.
As for me and my type? Well, I like to think that I make new things as a business leader in corporate America. But what I do most of the time is make decisions. Decisions that yield outcomes that the world has already defined for me: More customers. More revenue. More profit. If there were a sweet spot for a computer, its choosing an option that yields defined outcomes.
If I don’t start adding more value with the creative parts of my career, then I’m gone too.
Excited for the future yet?
This is probably a good time to bring up that the bar to replace a human isn’t that artificial intelligence needs to be perfect. It simply needs to be better than humans relative to investment. We’re there already for many, many jobs. And for every argument that human judgement and understanding yields an absolutely necessary ingredient to the work being done, there’s an equally compelling one that, for certain roles, the avoidance of cost and the variances of error, carelessness and even corruption, on the margin, makes the machine better for business.
If you love lubricated free market capitalism, then you love machines.
Which lands us on the most material reality of this discussion. Sometime in the future, around the time my children’s children are about to start their education and figure out what to do with their lives, we, America, as a free market, democratic society, are going to have to decide to either A) limit the advance and use of technology in our economy or B) establish a form of universal basic income funded by the corporations that profit from the elimination of the American work force.
Here’s a fun consideration about option “A”. China isn’t going to do it. And neither will North Korea. Or Russia. So if we do, we can wave at them kindly when they wander past us into the future of global leadership/domination.
We’re either ok with that. Or we’re not.
At least we’ll still be free…for a little while that is.
We’ve got a generation and a half before we get to answer a lousy question. What part of the American way gets eaten by the machine? Is it the free market competition that has driven us to where we are? Or is it the delicate dance of inequality and opportunity that’s delivered centuries of self-determination. Which means it’s probably time to start thinking about it now.
When you don’t like option A or B, it’s best to get to work on option C long before it’s time to pick one.
If a year from now this scary reality seems no closer, and you’re compelled to point to we technology alarmists and gloat, I’ll leave you with my favorite Bill Gates quote.
“We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten. Don’t let yourself be lulled into inaction.”
We’re standing on the tracks. We can either get off, get run over or find a way to jump on and ride the train that’s bearing down on us.
But one way or another, the train is coming.