Antibuzz has been running as a weekly column (more or less) for three years. We have decided to give Andrew a few weeks off and run some “best of” columns from the past. This column was published originally in August of 2010 and introduces a common Anti Buzz theme.
The Buzz: Computers are taking over the world! They are all smarter than us and we are doomed!
The Anti-Buzz: Computers are not smarter than us, and Skynet is not on the horizon.
Why: Just because a machine can do arithmetic in less than a nanosecond does not mean it is intelligent.
Let’s talk about Artificial Intelligence.
First, AI is a very ambiguous term. In Computer Science it is used to reference a good deal of topics, most of them very non-threatening to the continued existence of the human race. Conversely, in popular culture it almost strictly means the replication of human intelligence on a computer, and in the movies such technology seems to invariably lead to a machine revolt against its masters.
This pop culture view of AI is what I intend to discuss today. The other sub-categories will be discussed in the future, most especially “specialist AIs” which can and will have an impact in medicine, including dentistry. But that’s another day.
Today, I’m here to assure you that all those pop-culture fantasies of machine revolution are just that: fantasies.
Your Desktop Is Not Smarter Than You:
The first misconception to nip in the bud is the idea that intelligence means you can do math really fast, among other things. Rules of arithmetic are just that, rules. Addition and multiplication work that same way no matter what numbers you cram together. If you are slow to add numbers, it either means you forgot the rules, or you are out of practice. It does not mean you are stupid. Mathematical operations are inherently predictable, and thusly engineers eventually figured out how the process could be automated mechanically. Such arithmetic does not comprise “thoughts” floating around your computer’s head – it’s just a mechanical process.
I am reluctant to point fingers, but I do think it is something that generations older than mine are more guilty of: Math and certain other topics have an association with intelligence, and so therefore any person or thing that can perform well at those tasks must be intelligent itself. Further computers are themselves often the source of much confusion, especially to new comers. If it performs math quickly and is a device that you don’t understand, it is actually quite natural to assume some intelligence.
I know that we’re mostly joking when we say our computers are smarter than us, but this sort of misrepresentation of what intelligence really means is at the heart of the popular notion that we are always just a few years away from the Skynet doomsday scenario.
I’ll tell you what does require intelligence –
I’ll tell you what does require intelligence – recognizing that arithmetic can be a mechanical process. This required a great understanding of math on an abstract level, the proactive thinking required to want to create a new technology, and the ability to not only imagine something new, but to imagine how one might build it. None of these are things that any computer on this earth are capable of.
Deep Blue Beats Garry Kasparov:
But clearly playing chess requires intelligence, right? Certainly, but when the Deep Blue AI beat Garry Kasparov so many years ago, it wasn’t an AI in the Terminator–Cylon sense of the word. It was a feat of software engineering, no doubt, but it was just a very complex and intelligently planned set of instructions that happened to play a really good game of chess. This AI was capable of absolutely nothing else. It could beat you at chess, but it didn’t feel one way or another about it, and it didn’t go home at the end of the day to some loving family where it might talk about the chess games it played that day. It wasn’t bitter that its only function was this one specific game. It was just an algorithm that could play chess. Not a person. Not even close.
Like the computer itself, Deep Blue was of intelligent design. It was made by a number of intelligent people who could calculate ways in which solid chess strategy could be automated. If we ever make a computer program that is not intended specifically to play chess, and yet it can dynamically figure it out on its own, then yes, maybe the Cylon invasion will be imminent. But Deep Blue didn’t solve chess on its own, a bunch of humans told it how.
Maybe that’s the best explanation: Computers are abyssmally stupid, but they are tireless, fast, and very, very, very good at following directions.
The Appearance of Intelligence:
By playing a good game of chess, Deep Blue seemed intelligent. The appearance of intelligence is often a goal of AI researchers.
Imagine a robotic tour guide that responds to your questions in a human fashion, makes sense of your human speech, no-matter how idiosyncratic, is flexible and can begin discussion of a number of topics, is not boringly predictable in its responses, and otherwise does a very good job of behaving like a tour guide.
Such a thing is not too far beyond our capabilities now, but it would not be intelligent. Like math and chess, it would just be privy to a rather well designed set of rules for interacting with humans in a tour-guide-like fashion. Social interaction can not truly be broken down into rules like math and chess can, but it can be approximated with increasing efficacy as we venture forward.
So What IS Intelligence?:
If I truly attempted to answer that question, I would stray too far from this blog’s intent. Also, I am hardly equipped to answer the question with any authority. In previous drafts of this very article I went down that rabbit hole. This is not the place.
I think the best point to bring forward is simply this: Human intelligence tends to be far greater and more complex than a lot of people want to believe. Through all of history the trendiest attitude has been to assume the worst of humanity, from self-loathing Puritans to modern hippies, people have been grossly underestimating, well, people.
Conversely, at this point in time, computers are only machines, and in truth no more capable of independent thought than your lawnmower. They might be much more complex and flexible than your lawnmower, but just because Windows sometimes does things without your asking doesn’t mean that some sinister intelligence is evolving underneath. Computers do nothing without input, only ever receive input when they are told, are incapable of ignoring instructions in their programming, and so on. There is nothing about them that drives them to act on their own, nothing that allows them to resist their own coding.
In short: Skynet, with current machinery, is impossible.
If you’ve heard the phrase machine learning, its really just another approach to algorithm design. If I made a machine learning chess program, rather than giving a specific set of rules for playing chess, I would give it a specific set of rules for analyzing chess-related data and then building its own rules from that. Even so, these new computer-generated rules are only designed according to my own rules. Its just another layer of abstraction, a common theme in computer science.
However, this idea is at the forefront of all genuine attempts to recreate human intelligence. The most impressive research I have seen tends to focus on human psychology, and then attempts to emulate how humans seem to learn one specific type of thing. Very narrow, very small, but this is probably the best way to glimpse how Skynet will eventually be conceived. The thing to remember is that when that day finally does come, the new AI will, in a manner of speaking, be just a a baby. It might be an intelligent being, but it won’t be out-of-the-box confident, capable, or knowledgeable. It will grow up, develop its own personality and then … who knows. But it won’t be just an on-switch for Armageddon.
So, to wrap up this week I think the fair thing to do is ask how this sort of thing flies with the readership? It would be inappropriate to get so theoretical and philosophical every week, but I also don’t think it is meritless to address popular conceptions of computers from time to time. I also want to jump around and address a breadth of topics, rather than hammer on similar points for weeks on end before moving on and never returning.
Next week I will be severely practical to make up for this week. Until then, relish in the knowledge that not only are you smarter than your computer, but you are smarter than all of them combined.