AI has been predicted for decades. There's a reason why it doesn’t exist. Yet.
Episode Summary:
Artificial intelligence has been predicted in popular culture for over a century. The idea that human beings can make thinking machines is so popular, and so pervasive that we tend to call any sophisticated algorithm, intelligent. There is no artificial intelligence, at least not yet, it’s possible that there will never be. Coming to terms with this however, isn’t easy.
Access all episodes of End of the Line on Engineering TV along with all of our other series.
Transcript of this week’s show:
There are lots of lines that people tell us every day, but lies that we all embrace, except, even enjoy. So my favourites are “maintenance-free”, “guaranteed for life”, and of course, “no new taxes”. But the biggest of all, and the one that’s hot all over the media right now is “artificial intelligence”. I’ll say straight up, right now. There is no such thing as artificial intelligence. There never has been. It’s possible that there never will be. There are several reasons for this, but in my opinion most significant is simple: since we don’t know what human consciousness is, and therefore can’t clearly define human intelligence, we create a synthetic analog of it? But even if we could fully understand intelligence in the wetware, the notion that brute force algorithms running at light speed on silicon somehow magically attains this thing we call intelligence is obviously untrue. Computers can now routinely defeat chess masters at the game, but even those famous IBM monsters like Deep Blue didn’t think like human beings. The ability to iterate millions or even billions of possible outcomes, then choose the best path forward simply isn’t the way that intelligence operates. Intelligence operates subconsciously. We don’t know why 2+2 = 4, and we don’t count on our fingers to know it. It’s not just the rote memory of childhood arithmetic that makes us intelligent, it’s the ability to take concepts like that and use them to understand much more sophisticated things, things we have not learned through repetitive training. Manned spaceflight to the moon was developed using computers, but was also developed using these in the hands of skilled engineers, who thought in sophisticated ways, like the logarithms used by this slide rule. That’s actual intelligence.
Don’t get me wrong, machines are pseudo-intelligent and are essential. They can learn, in the sense that they can recognize an erroneous path and stop before they go too far down that road. But if you want to see what I’m talking about, take a look at the new and powerful tool called generative design. Want a lightweight, high-strength bracket? Let this new CAD technology iterate a billion or so designs, run the FEA and deliver an optimized solution. But it will do so always within the constraints of predetermined physics and available material properties. I once worked for a brilliant senior engineer, who once pointed to his very large Rolodex and said: “this is the key to engineering”. What he meant was that when he couldn’t imagine a solution to a problem, he could imagine that there was a business card of someone he had met in that Rolodex who had the necessary knowledge and creativity to help solve that problem. Today we tend to do this with a Google search, but I guarantee that the good stuff is only accessible by talking to a creative expert, not doing an automated literature search. Every well-designed and efficient algorithm these days is called artificial intelligence. It isn’t. When it stands next to me at the water cooler and says “who’s that guy we had that drink with at Aerodef last year? The guy from LockMart, he understands this stuff.” When the machine can do that, pipeline me directly to an innovative and creative solution to a problem, then establish a personal connection that will work to the advantage of both of us for years in the future, then I’ll call it artificial intelligence.
