Artificial Intelligence has been creeping up on us for years. Now it’s here.
My First Experience With Artificial Intelligence
Many years ago I wrote a computer program for the Coldwell Banker real estate group to predict the selling price for houses. When a broker works out the pricing for a home, they look over the “comps,” that is, the comparable selling prices of similar homes. The problem is that no two houses are exactly the same. Even if they were built identically, they might be in different subdivisions, or one is on a corner lot, or a zillion other factors. Time is also a factor as market conditions change. Over time, brokers develop enough experience to intuitively (hopefully) make a pretty good guess as to the eventual selling price.
However, there is a mathematical process called multivariable polynomial regression analysis that can make some sense of all the factors. To give you an idea of how it works, suppose the ONLY factor that determines a price is the square footage. From similar houses you find that a 2,000 sq. ft. home sold for $400,000 and a 3,000 sq. ft. one sold for $600,000. The house you are appraising is 2,500 sq. ft. so you can split the difference and get a likely selling price of $500,000. You could make up a graph of price versus square footage.
However, the square footage is just one of many factors, each of which adds (or subtracts) from the selling price. You are faced with questions such as:
- how much does a half bath add to the price?
- What about the width of the lot?
- Distance to schools?
- Subdivision?
- How about a swimming pool? Fireplace?
- Interest Rates?
The list can go on and on.
With only square footage as a factor and assuming it behaves linearly, you end up with a simple 2 dimensional graph. We were going to need over a hundred factors and a corresponding hundred dimensional graph. There is no way to visualize that, or really even to imagine it. However, all a computer sees is a bunch of numbers and it can handle it as well as it handles a Solitaire game. (If you are interested, drop me an e-mail and I’ll bore you to death with the actual mathematics.)
We spent six months loading up all the “comps” we could find, then tried it out on some new listings and also some of the old to see how well it predicted. It worked amazingly well! I don’t know if the successors to Coldwell Banker or other real estate groups are using anything like this, but they probably should be.
A Rude Awakening
But then, I took a look “under the hood” and started to get concerned, maybe even a little frightened. I’d find things like in one subdivision the AI would up the price by $1137 for a half bath, but in a different subdivision it might add only $42. I even found one case where a half bath dropped the price about a hundred bucks. WTF????
It dawned on me that, while it was getting the “right” answers, I didn’t have a clue how it was deriving them. The program had become a magic black box. I knew as much about its “internal reasoning” as I did the internal reasoning of my cat. It did a better job of predicting price than I could, a better job than most brokers, but it was getting there in a totally different manner than we humans were using.
I was concerned (and fascinated) so I started doing some research. My first stop for technical papers was from that “school in Massachusetts” that decent people never mention. Decent people being those to be found in Pasadena, California – with a few in La Cañada Flintridge – and mostly to be located between East Del Mar Blvd and East California Blvd north to south and between South Catalina Ave and South Hill Ave west to east. By the way, not to brag, but the “real school” hasn’t lost a football game since 1968!
(*Editor’s note – 1968 was the last time Cal Tech played a football game – gotta try to keep this joker honest. It ain’t easy!)
I read the papers, particularly those from Marvin Minsky and his collaborators at the unmentionable school, and found they were seeing the same startling results as I was. John McCarthy, who was at the “real school” for a while before he started slumming around, coined the term Artificial Intelligence in 1956.
Alan Turing (remember the movie “The Imitation Game?”) had proposed the “Turing Test” in 1950. In the test, an investigator converses remotely with either a person or a computer. If the investigator can’t tell the difference between the person’s and the computer’s responses, then the computer passes the Turing Test. It is displaying artificial intelligence at a level such that we can’t tell it from human intelligence.
We culturally have come to think of artificial intelligence as human intelligence that is being mimicked artificially. Alexa, Cortana, IBM’s Watson and others are creeping up on the Turing Test. Doug Lenat at CycCorp in Austin has for years been trying to program “common sense” into computer systems.
It all sound good, and progress is very promising. So what’s the worry?
The problem is that AI isn’t Artificial Intelligence. It is ALIEN INTELLIGENCE.
The “thought processes” behind my rudimentary program, or the “thought processes” behind Alexa, Cortana, Watson, etc. are nothing like human thought processes. Alexa’s “mind” bears as much resemblance to a human mind as would the mind of a three foot tall blob of gelatinous shellac that just landed its flying saucer on the White House south lawn. Maybe we could converse with an alien. It could pass the Turing Test perfectly, but what it thinks, what its motive are, what its intentions are, remain totally alien to us. And, we’d be just as alien to them.
Intel and the other big chip manufacturers are building dedicated AI chips, most of the Fortune crowd is adopting AI into their business practices (some much better than others,) and Alexa and the rest are becoming more immersed in our day to day lives. We are watching the dawn of silicon sapiens.
It’s a new species, a new intelligence, and a new very alien culture entering our society. Where it will lead, I can’t say, but the potential damage is enough to scare me, just as it was enough to scare Stephen Hawking. Perhaps it will lead to a human utopia where everyone leads the life of a billionaire, or maybe it will lead to a time when humans are only tolerated because of their skill with opposable thumbs.
We are becoming exponentially capable of programming new tasks for the silicon sapiens. I just hope someone is working on how to make them like us.
(Go Beavers!)
