Artificial intelligence is new, but is it really different?
Fear that artificial intelligence will replace humans and create a miserable and dystopian world has been a staple of science fiction for decades. But will it actually happen? The simple fact is, humans have been relying on algorithms running on computers for seven decades now, and although the code is created by a human, the answers that it generates are generally accepted as fact by people who don’t understand the code, or how it was generated.
This faith in technology is the cornerstone of modern society and from the perspective of the average user, including engineers, computer software already operates as a kind of artificial intelligence. What matters are the answers, and the questions—and until AI learns to read minds, that will always be where humans dominate.
Access all episodes of End of the Line on Engineering TV along with all of our other series.
* * *
Episode Transcript:
Are you afraid of that artificial intelligence will send the world into some kind of dystopian hellscape as seen in movies like the Terminator? Worried that artificial intelligence will run out of control? Concerned that it will eliminate everyone’s job and create mass poverty?
Well, that’s a popular storyline, but I’m more than a little skeptical that AI is the road to ruin.
Let’s do a little thought experiment. Imagine that you are given a simple task: drive to your local 7-Eleven for a Big Gulp. You know your current location, and you know the location of the 7-Eleven. What’s the most efficient way to get there?
If you’ve never been before, you could consult a map, then make a decision based on the shortest route length, or perhaps one that uses streets that you know to be relatively traffic free. If you were a computer algorithm doing the same task, you might virtually attempt every possible route to the store, measure the distances or the travel time, then choose the most optimal one. At a higher level of sophistication, used by some traffic apps like Waze, real-time data is fed from other vehicles already operating in the local environment, which flag the algorithm about routes that are clogged, suggesting a new optimal path.
But the algorithms operate the way a small child would operate: try everything, then learn by trial and error. The concept of artificial intelligence is that it would operate the way we humans do: look at an overview of the problem, then immediately select a perfect or near perfect solution, based on common sense and our existing knowledge of how traffic operates.
But from an end-user perspective, what difference does this make? An unintelligent algorithm that operates with hardware that can execute instructions so fast that billions of iterations can be worked through in milliseconds will deliver exactly the same outcome, and the way it was derived will be imperceptible to you or me.
In other words, artificial intelligence and relatively unsophisticated iterative problem-solving is indistinguishable if it delivers the right answer, quickly. But notice that the key factor isn’t how a computer generates an answer—it’s the context built into the question.
In my thought experiment, your goal isn’t to discover the fastest way to that 7-Eleven at 8 o’clock on a Thursday night. It’s to get a large soda. Systems like Amazon Alexa or Google Home are better at handling natural language queries. Ask it where you can get a large soda, and these systems will give you several options. But no matter how sophisticated the AI, they are constrained by the context built into the syntax of the query, which means humans are still invaluable.
In my thought experiment, if I asked my wife the quickest way to the 7-Eleven, her answer will likely be that I drink too much soda anyway, and that caffeine is a bad idea this late in the evening—because she knows why I’m going, without me telling her. That’s intelligence, and I don’t think I’m going to live long enough to see a machine that can do that.
If you’re an engineer, the lesson is simple: it isn’t about the answers, it’s really about the questions. Get the questions right, and it is relatively easy to get the answers, whether you’re plotting a spacecraft trajectory, or designing an electron microscope.
AI will be the best tool that engineers ever had, and better still, it isn’t going to replace anybody.
