I’ll say it right off the top. It isn’t gonna happen. Notice, I don’t say it couldn’t happen. I mean, Donald Trump got elected president, didn’t he? But Artificial Intelligence becoming smarter than humans and taking over the world? The odds are heavily against it.
And if it ever happens, I guarantee that somewhere in the background there will be a human directing things. Because the old saying is true. Guns don’t kill people. People kill people. Same with computers. It’s a matter of motivation.
But They’re So Smart Already
No, they aren’t. A chatbot is smart like a parrot. Parrots aren’t smart. We just think they’re smart because they mimic human speech. But they aren’t talking.
Let’s take “Polly want a cracker,” for instance. The parrot has learned that if he makes that noise, he gets a cracker. You think he’s speaking to you.
But what if you want him to learn to refuse the cracker? If the parrot was truly speaking, you could teach it that adding “doesn’t” reverses the meaning of the sentence. But that doesn’t work. In order to get the parrot to learn “Polly doesn’t want a cracker” you have to teach the whole noise (sentence) from scratch.
Computers are like that. When you ask a chatbot a question, the computer tries to pair that question with all the appropriate responses to that question fed into its memory in the past. It picks the one most used and spits it out. The computer has no idea how appropriate, how compassionate, how true or how evil that answer is. It just gives the answer statistics tells it to.
Parole System AI
So, we must be careful how we use AI. One judicial system in the US tried to use AI to predict how likely each applicant was to skip bail. They entered a bunch of data about each person: family situation, residence, past legal history, etc. and the machine predicted the results. It was a completely dispassionate system without all the biases of human judges and parole officers. And it worked. Bail jumping statistics dropped.
And then they had to shut it down. Why?
The whole concept of the judicial system is that people can be influenced to change their behaviour. If their bail application is dependent on their past, that can’t happen. AI can’t predict what any individual will do, only the odds of what an individual like that will do. Only humans can make a decision like that about another human.
Plus, there was no way to appeal. The decision to deny bail was based on statistics. A lawyer couldn’t say, “But you failed to consider this and that, and have you thought of this other factor?” The only answer to an appeal is “Your client is the wrong type of person.” As if we didn’t know that already. Score one for humanity.
Computers Taking Jobs
That’s a definite “yes.” History has shown us that if a machine can do a job faster, better, and cheaper (well, pick any two) than a human, sooner or later the pressures of the market are going to make it happen. Ask any factory worker.
Self Checkout
It’s gonna happen. All you Luddites complaining about it on social media, there’s no point. It reminds me of the old lady interviewed on the news in England in 1970 when the New Pence coins were introduced. Her solution? “Well, I just won’t use them, will I?”
Good luck, lady.
The fact is, there are two functions of a checkout system. One records the purchases. The other helps the customers. Recording numbers is what computers do best. If you’ve ever used a chatbot on an internet site, you know how useless they are. So there needs to be someone at the checkout desk to help the people who really need help. The rest of us are quite happy to whip through the Self Checkout and be on our way.
And if you decide you’re going to go to the human checkout out of pure pigheadedness, keep in mind that you’re using the live clerk’s valuable time, Shoppers in the line behind you need human service, and you’re holding those people up as well.
So, you young people who are choosing your careers; there’s one more prediction you have to make. If you’re going into a trade, consider how Artificial Intelligence will affect that job.
Computerized Mediocrity
One light gleams from the end of this tunnel. All the AIs are programmed with everything that has been created relative to their job. They rarely think up something new that works. So, while a chatbot can write an essay on Proportional Representation in voting, it will be an average essay. And the more computers write essays on that topic, the more average the pool of essays the AIs learn from will get. So, if you’re only average at a job, your occupation is in peril. If you have a creative bone in your little finger, you have an edge on the AIs.
Last and Happiest Thought: Computer Density
An evil super-computer is not going to arise out of nowhere. By the time technical progress has reached the level where such a computer is possible, there will be millions of computers in the world who are on our side. Your evil AI doesn’t have to be smarter than humans; it has to outwit the whole tech interface.
The Bottom Line
The present-day threat from computers is that AIs are smart enough to be used by average criminals to commit crimes. Tech whizzes have known for years how to put a person into a video clip, but now any fool can do it. The occupation of forger has been around for centuries, but now any fool can cut and paste a convincing document of any sort. Watching the evolution of what used to be paper money gives us an idea how that battle is going.
So, let the evil AI fade into the problems of the future. We need to concentrate on the human criminals.