“Don’t look back. Something might be gaining on you.”
— Satchel Paige
Perhaps the tech reporters who have been praising innovations that allow robots to replace human workers should ignore that advice and look over their own shoulders. AI evolution is moving us toward the rise of the machines faster than they anticipate.
The Robot Reporter
The Chinese have invented an AI that can write a 300-character news article in just one second. The robot reporter, named Xiao Nan, created the article for the Chinese media outlet Southern Metropolis Daily. Its developer, Wan Xiaojun, is a professor at Peking University who is working on developing several AI machines.
He sees only the benefits, of course. “When compared with the staff reporters, Xiao Nan has a stronger ability to analyze data and is quicker at writing stories.”
Human reporters should not worry about AI evolution replacing them however, because Prof. Wan reassures us that,
“Robots are still unable to conduct face-to-face interviews and respond intuitively with follow-up questions. They also do not have the ability to select the news angle from an interview or conversation.”
We keep hearing this kind of reassurance from experts who tell us that AI won’t be taking over the world any time soon because they can’t really think for themselves or act independently. No Skynet tomorrow, nope. Just can’t happen. AIs are just not that smart.
AI Evolution Begins
This article, “The Mind-Blowing AI Announcement by Google That You Probably Missed,” by Creative Technology Gil Fewster casts a spotlight on exactly how fast artificial intelligences are learning and developing independent of human direction. Mr. Fewster examines a December 2016 Google article published under a title only a techie could love: “ Zero-Shot Translation with Google’s Multilingual Neural Machine Translation System.”
It seems that Google has developed a new language translation tool, the Google Neural Machine Translation System. That doesn’t sound too threatening—who hasn’t wished for a device that would make translating from one language to another simpler and easier? It’s the kind of thing that helps humans communicate with one another, right?
The Smarter AI
Well, the GNMT kind of blasted right past that. The AI smartened up and figured out that the best and fastest way to accomplish translation was to create its own language. No human directed it to do this. No human programmed it to do this. The GNMT learned from the people who used it and made, “educated guesses about the content, tone and meaning of phrases based on the context of other words and phrases around them.”
The GNMT did all this in a matter of weeks.
So how long do you think it would take an AI that can learn this quickly to figure out more complex tasks, like interviewing a human being or selecting a news angle? How fast do you think it could go from Prof. Wan’s assurance that, “robots will be able to act as a supplement, helping newspapers and related media, as well as editors and reporters.”
Sure they will. All we need is some professor or researcher with more arrogance than foresight or more intellect than common sense to program ambition into an AI. Or self reliance. Or ruthless efficiency.
The Rise of the Machines
Well they’re starting small. Sean Martin in Express tells us about, “Robots Being Developed That Have a ‘Brain’ and Can Learn New Things Like a Human Child.” This time the genius developers are building a robot that can “mimic the human learning experience,” by learning from nothing as a human child would. The Italian firm Goal Robots hopes to have their first learnable robot within four years.
Essentially, they are programming curiosity into their robot with the aim of “finding something surprising.” Then the robot will figure out how it works or how to make it work.
Project Coordinator Gianluca Baldassare of the Institute of Cognitive Sciences and Technologies of the National Research Council (CNR-Istc), says, “We do not give them the goals from outside, they have to choose them themselves.”
Great. How many of us have lived with two year olds? Do we always approve of the goals they set for themselves? Or the way they go about reaching them? Do we really need a robot child that has just learned to say “No!”
The Bigger Question
What is the point of all this AI evolution? Why do we need a robot reporter when we have a multitude of human ones? Why do we need a robot that can learn like a child when we have our own children? Do we really need an AI that makes binary decisions about efficiency? Do we need an AI that gets creative all by itself when no one is watching?
And then we have the bigger question: How do these AI developments help humanity? Google and the other genius developers don’t ever seem to address it. They pursue technology for its own sake with seemingly little thought for the potential downside.
The darker issue is, as some of us have said all along, that these robots won’t help humanity at all. Instead they will hurt it in ways that we have only so far dealt with in realm of science fiction, which most people don’t read. Will we end up with C3PO, R2D2 and BB8 or Skynet and Judgment Day?
And do we want to find out the hard way?