How AI Works, Part 1: What Is Intelligence?
In this series of posts, we’re going to explore some of the concepts behind how AI works, so you can see how I approached the background for introducing and discussing AI in my novels. These posts aren’t about applications, or how people are using AI today. We already have a pretty good idea of that from our day-to-day use of it. What I’ll explore is what’s happening underneath and behind it all: what is intelligence?
To begin with, I don’t think these AI models are really “artificial intelligence.” But neither are they simply “really good computer programs.” They’re something in between. Things are also changing faster on the AI front than with any other technology we’ve ever created. So while we rush headlong into what I think of as the “capability gap,” between what AI can do and our ability to keep pace with it, let’s look at how we got here.
Everyone Already Has AI
Almost everyone with a computer or smartphone now has some form of what we call AI today. Most ordinary search engine requests are now being supplemented by AI systems in the background, with the responses we get becoming ever more sophisticated and increasingly tailored to our catalogued and stored online behaviors and preferences.
Much of the information used to build these “behind the scenes” profiles of who we each are comes from things we do online, often without most of us having any real understanding of just how much is being collected. Yes, we may have accepted those pages and pages of terms and conditions. But accepting them isn’t necessarily the same thing as understanding what we agreed to.
What We’re Actually Trying to Build
Let’s start by thinking about AI as a field of computer science concerned with getting computer programs, and by extension physical systems controlled by computers, to do things that would normally require human intelligence.
Thought of this way, we’ve actually been working on this problem for as long as computers have been around. Researchers have been asking what lay beyond routine processing, something more than a complicated calculator or a chess-playing program. How do we get a computer to do something genuinely useful?
What separates humans from the AI systems we’ve built so far isn’t simply the ability to process information. It’s our ability to understand the world around us, to reason about it, to apply “common sense,” to understand and use language, and to express complex ideas. We also have vision. We can translate what we see into a feedback loop: we do something, we observe the outcome, and we understand cause and effect.
Can Intelligence Be Just That: Mechanical or Biological?
These abilities allow us to learn, and to use something else we call intuition. They allow us to move through the world and understand the consequences of our actions. Add them all together, and you have what we unconsciously think of as an intelligent being, or in the terms we actually use, a person.
So the question we’re rapidly approaching is this: can intelligence itself be artificial? This is also the realm of science fiction writing. For my techno-thrillers, I tend to want to stick with what’s possible, or perhaps “nearby fiction,” things we’re clearly very close to already.
And I’m not even sure “artificial” is the right term for this. We’re an “evolved” intelligence, after all, which really just means our intelligence is biological. Our neural circuits, our ability to store memories, to see, to act and react, are all built into a biological system. What we’re really asking is whether the capabilities of such a system can be built into hardware instead, using memory chips, processors, and all the other components that make up a computer. And I suppose the more interesting question is, what would it actually take?
But our intelligence is more than this too. It comes from experience, culture, and language, from talking with other people, reading, and creating, and from finding ways to pass what we learn from one generation to the next. Each step lets us capture and pass on more information, faster. In that respect, the advent of AI represents another shift in the same long process, a “new gear” in how fast we can process information and create knowledge.
Processing Is Not the Same as Understanding
And now there’s something genuinely new: computers can extract patterns from enormous quantities of data, identify relationships humans haven’t previously seen, and produce responses that look remarkably like understanding.
What makes this prospect exciting and scary in equal measure is its sheer scale and speed. But there’s an important distinction here. Processing information isn’t necessarily the same thing as understanding it.
A computer can store millions of books without understanding a single sentence. And an AI model can produce an astonishingly convincing response without necessarily understanding what it’s said. We’ve all seen examples of this already, when an AI gives us a response that sounds perfectly plausible at first glance but on closer inspection simply makes no sense.
That distinction is going to become increasingly important as we look at where AI is heading. Because if a machine can process information, identify patterns, solve problems, and produce responses that appear intelligent, but doesn’t actually understand what it’s doing, then what exactly are we looking at? We all remember the kid at school who was incredibly “smart” but had no common sense or social skills. So is any of this really intelligence?
Or is it something entirely different, something that simply produces many of the results we think of as intelligence? And if we accept that possibility, a more interesting question follows: can different AI models have different levels of intelligence, capability, and even social skills, just as people do? That’s definitely an area I’m exploring in my next techno-thriller.
In the coming weeks, I’ll look at what’s actually happening inside these systems, what some of the implications might be, and how I think about AI when I bring these ideas into my fiction. Because before we can decide whether we’ve created artificial intelligence, perhaps we first need to decide what we actually mean by intelligence.