With enough data, you can fake intelligence. I’ll explain that in a moment, but first, let’s talk about intelligence. Are bacteria intelligent? I don’t think so. Are dogs? Maybe. Are dolphins and whales? Almost certainly. Intelligence is a bit like pornography: You know it when you see it (a phrase made famous by Supreme Court Justice Potter Stewart’s obscenity ruling in 1964).
Alan Turing, one of the pioneers of modern computing, proposed what’s become known as the Turing test as a means of deciding whether a machine can think. A human is placed in a room and can ask questions of two unseen players, A and B, via text chat. The object is to decide, after five minutes of texting with each player, whether A is the computer and B the human, or vice-versa. Of course, beating the game really only requires a sophisticated human conversation simulator, including the ability to delay answers (to simulate human thinking time) and make typing mistakes (for obvious reasons). But it doesn’t really say anything about intelligence.
More recently, the IBM computer program named Watson successfully beat Ken Jennings and Brad Ruttner at “Jeopardy!”. Jennings and Ruttner are arguably the two strongest “Jeopardy!” players in the world, and although Watson was better at “buzzing in” than the human players, it still had to get the right answer most of the time. The software that underlies Watson is now being used to help nurses make decisions about lung cancer treatment at Sloan-Kettering Cancer Center in New York. Is Watson intelligent? Probably not, but that doesn’t mean it can’t outdo humans at what have been traditional bastions of human intelligence. There’s also Siri, the helpful “intelligence” inside Apple iPhones, which attempts to provide useful responses to spoken questions and requests. Sometimes it even works.
Jeff Hawkins, who invented the Palm Pilot and later wrote an interesting book called On Intelligence, believes that our brains are basically pattern-matching engines. Matching sensory inputs to existing patterns in our brains lets us make predictions about what will happen in the future. If the predictions are correct, they create new patterns. His book is well worth reading (and I’ve recommended it here before). Numenta, the company he founded to explore and commercialize this technology, recently changed its name to Grok. To quote its website, “Grok software finds complex patterns in data streams and generates actionable predictions in real time.” Pretty heady stuff, but also incredibly valuable to lots of businesses if it works as advertised.
Artificial Intelligence (AI), the branch of computer science that deals with “thinking machines,” covers a lot of ground, which reflects the fact that “intelligence” isn’t just a single, well-defined thing. Even within the AI research community, there’s a distinction between “weak” AI, which accomplishes specific problem solving tasks (like Watson), and the elusive, “strong” AI, which essentially simulates human behavior: It sees; it learns; it makes decisions; it manipulates its environment; it speaks; and so on.
In 1994, I shared a dinner table with James Cannavino, who at the time was head of IBM’s personal computer division. He was a raconteur of sorts and proposed that someday in the not-distant future, we’d be able to simulate all the chemical reactions that take place in the human brain. And if you can do that, he said, it seems reasonable to assume that the simulation would act like a human brain. He then applied Moore’s Law to the whole concoction to imagine a human brain operating and evolving at rates that are orders of magnitude above the rate we see in the physical world. While there are some leaps of faith involved, it’s perfectly imaginable that in tens (whoa!) or hundreds (who cares?) of years, we’ll be able to simulate the human brain (and lots of other things) at the level of atoms and molecules. The biggest obstacle, in fact, is simply our understanding of biology. And if you listen to people like Ray Kurzweil, the Singularity—when intelligence surpasses the limits of human biology—is less than 50 years away.
Me? I’ll be happy with a driverless car in the next decade or so. But it’s good to keep in mind that we may not be at the top of the intelligence heap forever.
I was reminded of my dinner with Cannavino while reading a great book this summer, Avogadro Corp., subtitled "The Singulartity Is Closer than It Appears” ($2.99 for the Kindle version on Amazon). In the story, Avogadro Corp. is the largest Internet service provider in the world and handles email for almost everyone via its AVoMail program (can you say GMail? I knew you could). A clever programmer enhances the email program to ever-so-slightly rephrase your messages to make them more effective (in other words, more persuasive). It does this by looking at similar emails sent to the same (or similar) recipients that generated the desired outcome (a raise, a project approval and so on). And, as they say, hilarity ensues…or not.
But the basic idea (shared by Hawkin’s Grok) is that, if we can determine patterns in data (like, what makes for a persuasive email) and take action based on those patterns (like changing a word or a phrase), the result looks an awful lot like intelligence. And if a program has the ability to send those persuasive emails on its own, well, anything can happen. It’s a fun summer read, and I highly recommend it. In real life, the author (William Hertling) is a programmer for Hewlett-Packard. He’s followed Avogadro Corp. with AI Apocalypse, which considers machine intelligences that aren’t like humans ($3.99 for Kindle), and the soon-to-be-released The Last Firewall. I’ve read AI Apocalypse and found the writing a little weaker than in the first book, but still full of interesting ideas. I’ll be buying The Last Firewall as soon as it’s available.
There are already organizations with access to your email and tremendous computing power (hello, NSA!). Some are even well meaning (Google’s unofficial motto is “Don’t Be Evil”). The reason Avogadro Corp. is so popular is that its premise is entirely plausible. Read it for yourself and see.
