Remember when Deep Blue beat world champion Garry Kasparov at chess back in 1997?
Although Deep Blue won that six-game match, many AI commentators pointed out that the accomplishment was focused on a very narrow cognitive area – understanding the rules of chess, and selecting the best move to play within a specified time format.
While the accomplishment was significant, it was still more about programming and CPU speed than about anything we would call “machine intelligence”.
Now 14 years later, IBM is back. This time it’s all about Watson, a super-computer that will compete on Jeopardy against two of the all-time champions of the well-known quiz show.
“So what?”, I hear you say, but actually you should expect to see some mind-boggling technology at play here. Voice recognition, natural language semantic analysis, heuristic search algorithms… Just thinking about how to program all of that – and get a response within two or three seconds – is enough to make anyone’s head hurt.
In order to win Jeopardy (or even participate), Watson needs to handle not only a vast range of subjects, but also the cognitive challenges involved in understanding and analyzing Jeopardy clues (not to mention Alex Trebek’s sense of humor). These usually require some level of linguistic intuition and cultural awareness, in addition to encyclopedic knowledge.
So although it is still about programming and CPU speed, this time around the computer’s performance is likely to cross the line into a domain that most of us would consider “intelligent” in some sense.
Sceptical? Then think about the original test of machine intelligence proposed by Alan Turing – “A human judge engages in a natural language conversation with one human and one machine, each of which tries to appear human. All participants are separated from one another. If the judge cannot reliably tell the machine from the human, the machine is said to have passed the test.” [source Wikipedia]
Of course, the Turing test raises a host of questions, among which my favorite is whether the ability to behave indistinguishably from a human is, in itself, a definition of intelligence…
But I digress. Regardless of how one chooses to define intelligence (or lack thereof), it will be interesting to see how Watson performs tonight and the following two days. Make sure to watch these episodes of Jeopardy, if you are even vaguely interested about how advanced computer science has really become…
On Sunday Jan 30th I ran the ING Miami Marathon, my second-ever marathon. My finish time was 3:53:49. This was a PB, or “Personal Best” – not just on race time, but also for several other reasons. I broke the 4-hour barrier; I maintained a relatively steady pace throughout the race; I completed my longest race in Vibram FiveFingers; and I improved my previous marathon time by 45 minutes.
My main objective was to break 4 hours. To accomplish this, you have to run an average pace of 9:09 per mile. On shorter runs this is “conversational pace”, meaning you can run at this speed and talk comfortably at the same time. After 15 miles or so, however, it becomes very difficult to talk, or think, or in fact do anything other than just run. And after 20 miles, even your running starts to fall apart at the seams…
Because the last hour is the hardest, my race strategy this time around was to put as much time as possible “in the bank” up front, while making sure to keep enough energy to finish strong. My friend and training partner Richard’s goal was also to break the 4-hour barrier. Our pre-race plan was to stick together as long as possible and “carry” each other through the tough moments of the race, since this team dynamic had worked well on our long training runs.
If you are interested in the mile-by-mile account of my race, read on! I’m really writing this for myself so that I can refer back to it in the future and learn from my experience. But if you want to know how it feels to run a marathon, this is as good a place as any to start. Of course it’s long – but every minute you spend reading is twenty or thirty minutes that I spent running, so if I can do it, then so can you! Click here to keep reading
Finally after a “drought” of several months I pulled off a solid win as black using the Dutch Defense.
I favor the Dutch – and the Bird (1. f4) when playing white – because these systems make for dynamic play with several centers of activity. However, I don’t get to use the Dutch very often, because not many people open with 1. d4.
In this game I feel I gained the upper hand early. White’s moves 5. e4 and 9. dxe4 gave me the idea of bracketing the white king via the open d and f files.
Once white exchanged his knight for my bishop (11. Nxe7+) I felt the game was going my way. I like to keep my knights as long as possible in the Dutch/Bird systems. The pawn structures tend to hang around for a long time, blocking the diagonals and cramping the bishops. So having a knight against a bishop often proves to be a significant advantage.
12. … Be6 and 13. … Na6 connect my rooks, and on the next move I take ownership of the d-file, preparing to mount my assault up the center.
White realizes it’s high time to castle, but some diversionary attacks on his queen give me tempo to bring up the artillery and block his king in the center of the board. The game is not won at this point, but I certainly have a strong positional advantage…
I could not have anticipated white’s blunder 18. Nc3?? – but even without that, I think the game would have gone my way had we played it out. For example, 18. Qxd4 Nc2+ 19. Nxd4 definitively removes white’s castling option, and leaves his king rather exposed in the center.
It’s quite satisfying to win a game like this against a player rated 100 points higher!