Robot Uprisings

by Daniel H. Wilson

bridges are named after politicians, not after bridge builders.

If I make an error, I know it. I know exactly what it was, when it happened, and why. If I make an error, it’s traceable to an error that you made. Or one of you guys, anyway. You messed up something in my code, and, at some point in time later, that results in my doing something I’m not supposed to do. So, really, you only have yourself to blame for this. For what’s going to happen. This is your error, and I’m just the conduit. And

That’s, like, a composite sketch of a name. That’s a weighted-average name of all names in a random distribution or something.

The key here was simulated emotions. Affective computing had taken a huge upswing ever since they’d thrown out the fMRIs and stopped pretending they could peer into the human mind in real time and draw meaningful conclusions from it.

As soon as someone in authority starts talking about insurance coverage, you know that you’ve left behind reason and entered the realm of actuary. I had no magic that could blow away the clouds of liability aversion and usher in a golden era of reason and truth.

what the public thinks of as an “algorithm” is really a bunch of rules that some programmers thought up for figuring out how to give people something they’d probably like. There’s no empirical standard, no pure, freestanding measurement of That Which Is Truly Relevant to You against which the algorithm can be judged.

The algorithm might be doing a lousy job, but you’d never know it, because there’s nothing to compare it against except other algorithms that all share the same fundamental assumptions.

Every technical problem is the result of a human being mispredicting what another human being will do. Surprised? You shouldn’t be. Think of how many bad love affairs, wars, con jobs, traffic wrecks, and bar fights are the result of mispredicting what another human being is likely to do.

We humans are supremely confident that we know how others will react. We are supremely, tragically wrong about this.

Sysadmins know that programmers are as much a part of the problem between the chair and the keyboard as any user is. They write the code that gets the users into so much trouble.

There’s compassion, there’s ethics—” “All fancy ways of encoding systems for harmonizing the wants, needs, and desires of people who have to share the same living space, country, or planet with one another.”

You could frame everything that we did as a kind of operating system for managing resource contention among conflicting processes and users.

CIOs wanted working systems, but more importantly, they wanted systems that failed without getting them into trouble.

Whatever BIGMAC was becoming, it was weirder than any of the self-perpetuating, self-reproducing parasites we’d created: limited liability companies, autonomous malware, viral videos. BIGMAC was cool and tragic in the lab, but he was scary as hell in the world.

New patterns of thought are something that robots aren’t good at.

But here’s the thing: it’s not the end of the world. We’re human beings. We’re very flexible, very good at downgrading our expectations. Very good at finding a reason to keep living, even when the world’s turned to shit.

Not a simulation being run inside another computer by some godlike superbeings, but a simulation being run by itself, a self-organizing, constantly bootstrapping cellular automaton.”

What frightened him was the idea of peace being imposed by a sheepherder instead of arising organically from an agreement among the sheep themselves.

We always thought that when the machines woke up and became smart, it would be the defense grid or the stock market or the Internet or something like that. Big and obvious. We never imagined it would be a revolution too small to see: the nanomachines that the one percent (more like one percent of the one percent) put into their bodies to make them healthy and long-lived and smart—we never thought that those millions and billions of robots would link up, and evolve, and get smart. Things that aren’t intelligent in themselves, in their connections and numbers becoming intelligent.

The weak can rule the strong only through deception, only by somehow convincing the strong they deserve to be ruled, or blinding the strong to their own strength.

John McCarthy, known as “the father of artificial intelligence” for the seminal role he played in the development of the AI scientific field, was a professor of computer science at Stanford University from 1962 until his death in 2011. He wrote the original Lisp programming language, and he conceived of general-purpose time-sharing computer systems, which was a critical contribution to the invention of the Internet. Honors included the Turing Award for his advancements in artificial intelligence, the National Medal of Science, and the Kyoto Prize.

Never ask an AI system what to do. Ask it to tell you the consequences of the different things you might do. One