Tuesday, February 07, 2017

Artificial Intelligence?


One of the classic themes of science fiction is that of machines turning on their human creators - the ultimate expression of the Law of Unintended Consequences*. The franchise almost everyone knows is "The Terminator," which was based on D.F. Jones' 1976 novel "Colossus," which in turn was derived from a terrifying 1967 story by Harlan Ellison called "I Have No Mouth and I Must Scream." You could go even farther back, to a 1944 short story by Theodore Sturgeon titled Killdozer!.

This leads me to the discussion of an interesting, if disconcerting article by Steven Overly in yesterday's Washington Post: The Great Artificial Intelligence Gamble That Paid Off. It tells the story of a recent experiment conducted by Carnegie Mellon University to develop an artificial intelligence program that could play poker and beat human players - a similar effort to the one that developed "Deep Blue," the IBM computer that was able to beat reigning chess world champion Garry Kasparov in 1997. The poker program was able to defeat four professional players in one-on-one competition, winning more than $1.7 million.

The development of the poker-winning program was a significant milestone in the development of artificial intelligence (AI) because it showed that a computer can make decisions based on limited information and accumulated experience, the same way that humans do. It proved that AI could succeed in one-on-one competitions, although it was less effective when dealing with multiple players. As Mr Overly points out in his article, poker is a zero-sum game that operates within a well-defined set of rules and hierarchies of outcome, whereas decisions made in the real world “... often require compromise**, and whether one outcome is more or less desirable than another may depend entirely on the player’s perspective***.”

So, Dear Readers, what does this infer for our future relationship with machines? I, for one, am not ready for a machine that's smarter and makes better decisions than I do ... although most electric shoe buffers already fit into that category. Nor do I think I'm ready to hop into a self-driving car that relies on GPS and AI decision making on the rules of the road. While the future imagined by Theodore Sturgeon, D.F. Jones, Harlan Ellison, and the Terminator films may be overly dystopian, it isn't beyond the bounds of possibility. I think we need to be very careful about the way we integrate the use of AI in our decision making.


But then again, how bad can it be? We do a poor enough job relying on real intelligence, as evidenced by the outcome of the 2016 election.

Have a good day. Be more intelligent than your refrigerator. More thoughts tomorrow.

Bilbo

* The law applies equally as well to politics as to science fiction, as Trump supporters are slowly finding out. 

** Although real people aren't very good at compromise any more, anyhow.

*** This is also true, as we observe the radical differences in how "the people" view the actions of the Trump administration based on their political and economic perspectives.

4 comments:

John Hill said...

At least AI offers some intelligence. That might be a better government than our current Congress.
AI & alternative facts ... what could possibly go wrong?

Mike said...

I had what I thought was a fairly good handle on windows XP. Looking through many options on my new windows 10 has left me scratching my head on what they are even talking about. I think I've been taken over.

eViL pOp TaRt said...

AI is better than no intelligence at all. I wonder if intuition-based thinking serves a purpose too. Relying only on logic systems might provide a built-in flaw.

allenwoodhaven said...

Great cartoon! John, Mike, and pOp TaRt all make good points. We shouldn't move too quickly with AI.

How bad can it be? As Trump said, in his effort to be inclusive during the campaign, "What have you got to lose?".
AI could lead us in a direction we'd be better off avoiding, even if it's not as bad as The Terminator.