Willisms
Navigation
Buy WILLisms

XML Feed
rss-icon.gif Feedburner RSS
WILLisms.com on Twitter






Featured Entries

The Babe Theory Of Political Movements.
Mar. 21, 2005 11:50 AM

Iran's Sham Election In Houston.
June 20, 2005 5:36 AM

Yes, Kanye, Bush Does Care.
Oct. 31, 2005 12:41 AM

Health Care vs. Wealth Care.
Nov. 23, 2005 3:28 PM

Americans Voting With Their Feet.
Nov. 30, 2005 1:33 PM

Idea Majorities Do Matter.
May 12, 2006 6:15 PM

Twilight Zone Economics.
Oct. 17, 2006 12:30 AM

The "Shrinking" Middle Class.
Dec. 13, 2006 1:01 PM

From Ashes, GOP Opportunities.
Dec. 18, 2006 6:37 PM

Battle Between Entitlements & Pork.
Dec. 21, 2006 12:31 PM

Let Economic Freedom Reign.
Dec. 22, 2006 10:22 PM

Biggest Health Care Moment In Decades.
July 25, 2007 4:32 PM

Unions Antithetical to Liberty.
May 28, 2008 11:12 PM

Right To Work States Rock.
June 9, 2008 12:25 PM



Donate





Links









Search



Archives

March 2011
February 2011
January 2011
December 2010
November 2010
October 2010
September 2010
August 2010
July 2010
June 2010
May 2010
April 2010
March 2010
February 2010
January 2010
December 2009
November 2009
October 2009
September 2009
August 2009
July 2009
June 2009
May 2009
April 2009
March 2009
February 2009
January 2009
November 2008
October 2008
September 2008
August 2008
July 2008
June 2008
May 2008
April 2008
March 2008
February 2008
January 2008
December 2007
November 2007
October 2007
September 2007
August 2007
July 2007
June 2007
May 2007
April 2007
March 2007
February 2007
January 2007
December 2006
November 2006
October 2006
September 2006
August 2006
July 2006
June 2006
May 2006
April 2006
March 2006
February 2006
January 2006
December 2005
November 2005
October 2005
September 2005
August 2005
July 2005
June 2005
May 2005
April 2005
March 2005
February 2005
January 2005
December 2004




Social Security Reform Thursday.
January 29, 2008

Caption Contest Archive
Jan. 21, 2009

The Carnival Of Classiness.
Mar. 14, 2006

Quotational Therapy: Obama.
Apr. 4, 2008

Mainstream Melee: Wolfowitz.
May 19, 2007

Pundit Roundtable: Leaks.
July 9, 2006

A WILLisms.com(ic), by Ken McCracken
July 14, 2006




Credits

Powered by Movable Type 3.17
Site Design by Sekimori




WILLisms.com January 2009 Book of the Month (certified classy):











The WILLisms.com Gift Shop: Support This Site

giftshopbanner.gif











This Week's Carnival of Revolutions: carnivalbutton.gif



Carnival Home Base: homebase.gif

























Willisms

« January 2011 | WILLisms.com | March 2011 »

Jeopardy's Watson & Paul Bunyan.

SPOILER ALERT

Jeopardy geniuses Ken Jennings and Brad Rutter were absolutely destroyed by IBM's Watson this week in a three-episode contest, marking an important symbolic step forward for artificial intelligence.

Watson was not without flaws. In Final Jeopardy on the second day of three, the category was U.S. Cities, and the answer was, "Its largest airport was named for a World War II hero; its second largest, for a World War II battle." Obviously, Chicago is the question (O'Hare and Midway). Watson put "Toronto."

Dumb.

Also, I couldn't help but think that Watson had an unnatural advantage over the other contestants. Because Watson received the text of the question as soon as Jeopardy host Alex Trebek began reading it, the computer had a slight but important (and unfair) advantage over the humans in terms of beginning to search for and process the possible question. The humans must read the answer from across the room and/or listen to Alex read the answer from across the room.

Watson should have had to read the answer and/or listen to it from across the room. Timing and the stress of paying close attention is an important part of the game. It just is.

What ended up happening was that Watson simply was able to ring in before the other contestants (who likely knew the vast majority of the questions), time and time again. Bring back a supercomputer that has to scan across the room and find the clue on the board, and possibly even listen, and then we're talking. Part of how Jeopardy works is that you must have perfect timing on your buzz-in. You aren't allowed to click your buzzer until Alex has finished reading the question. But for Watson, he didn't have to listen and wait for Alex. He had some other way of knowing when to buzz in, in order to lock out other contestants. Another key advantage.

Moreover, Watson also struggled in a category about months of the year, and in a category about decades (even giving a complete non-sequitur answer at one point). He "learned" the correct responses after they were given, because Jeopardy producers allowed Watson's engineers to feed him that data after each clue, but to truly be intelligent, Watson should have had to literally listen to the correct/incorrect responses from the other contestants and enter them into his database himself. That's what humans have to do. Watson should have had to do the same thing. Instead, the computer received an advantage the humans didn't have.

Still, Watson was able to answer relatively complex questions fairly well, which, in conjunction with a human user, could be incredibly beneficial to individuals and society. The implications of computers that get nuanced language are enormous, just as steam engine technology was an enormous step forward for mankind.

In the same way Ken Jennings and Brad Rutter dominated at Jeopardy, the mythological Paul Bunyan was the best among men at many things involving strength and stamina. He could cut more trees than entire armies of men. But along came technology, accessible to those armies of men, and suddenly Paul Bunyan was not such hot stuff:

I couldn't help feel this parallel, even more than when supercomputer Deep Blue became the chess champion of the world. Chess is basically a mathematical game with finite moves, so it wasn't really all that shocking that a computer could become better than a human at it. Not just better, but thousands or millions of times faster.

With Jeopardy, though, we're talking about unpredictable clues that could come from almost anywhere. That's what makes Watson's feat impressive, even if the computer did have some unfair advantages over the human players. Watson, even if he did cheat in terms of getting the answers first and knowing precisely when to buzz in, still knew the correct response nearly every time. Fifteen years ago, search engines on the internet would give wildly wrong answers. You'd type in "UT football record 1961" and it would give you some guy's geocities website with a few pictures of L.A. Raiders football merchandise. It could take several minutes, on up to an hour, to find information like that. Today, you generally get pretty close to what you're seeking, within seconds.

Watson wasn't connected to the internet, which makes the computer's feat even more impressive. If you can fit all of that knowledge into a server room, perhaps we'll be able to fit all of that knowledge into a desktop-sized box one day. Or a portable device. And unlike the Great Library of Alexandria, which burned and wiped out vast amounts of human knowledge in moments, our knowledge today is everywhere. It's in the cloud. It's on millions of devices, like Duck Tales' Junior Woodchuck Guidebook.

Today, if a library burns, we can rebuild. We have the technology. And Watson will assuredly change the way consumers want to interact with computers. We'll now want our computers to be smarter, more intuitive, and more interactive-- as "computer" was on the original Star Trek. We'll also demand that our computers proactively help us in our daily lives. We'll want computers to remind us that Spring is coming soon, so it might be time to top dress the lawn. We'll want computers to constantly scan our purchase preferences and let us know that a 5 gig thumb drive down at Fry's is running for 3 dollars while supplies last, or some designer ski gear we thought about buying a few months ago is now on sale for 35 dollars, down from 200 dollars. We'll want our refrigerators to not just sense that the milk is old and suggest that we go ourselves a new carton, but look at our calendar to see if we're likely going to be eating at home, then order intricate ingredient lists for meals it has suggested based on other meals we have liked (and don't like), other meals our friends and family have liked, and what ingredients are fresh, inexpensive, and readily available at the moment.

We'll want to be able to ask computers not just factual questions, but ask computers to come up with complex theses on complex topics. Why did the global economy go to crap, and what can we do to get things back on track? How far along is Iran in developing nuclear weapons? Is global warming the next "bubble," scientifically-speaking? What, if anything, can we do to incentivize Americans not to let their kids become so obese-- or are we possibly overblowing the problem? What is the meaning of life? Create a plan for me that helps me healthily live to be 110 years old but doesn't make me give up Tex Mex. Give us a strategy for dealing with our budget problems over the long run without freaking out huge constituent groups politically or harming our economy or cherished freedoms.

Stuff like that. Computers will need to be able to craft solutions for all sorts of serious situations that require more than just searching for a singular correct response to a singular Jeopardy clue. They'll need to scour history, economics, politics, business, and technology, then think outside the box and tell us, "hey, humans, you know you could save 30 trillion dollars and likely grow the economy faster than your current trajectory, if you just did x, y, and z." And ultimately, it would still be real people making the calls, but computers could help us arrive at complex solutions for complex problems.

We're getting there. And Watson, for all of its limitations, is a big step forward on that.

UPDATE: The Washington Examiner posted this great 'toon along the lines of what I am talking about.

jeopardy-watson.gif

I think there will be a lot of uses for these more intuitive computers. Sports teams. Scientists. Real estate developers. Even chefs at restaurants. Everyone will benefit, but we still have a long way to go before we can fit Watson 2.0 into every hand-held device.

Posted by Will Franklin · 17 February 2011 11:59 AM · Comments (0)

Trivia Tidbit of the Day: Part 922 -- Energy Not Why Texas' Economy Is Outperforming Nation's.

Texas' Job Gains Diversified-

In the numerous articles that writers in Los Angeles and New York have written in recent weeks about how Texas is really not all that great economically despite the ridiculous volume of evidence saying otherwise, there have been references to the Texas oil industry, as if that small set of companies alone buffered our economy. If not for the energy sector, we'd be as bad off as anyone else.

I've been arguing this for nearly three years now, but it's not just oil:

Only 16,200 jobs out of 226,400 private jobs (around 7%) created in Texas over the past year have been in "Natural Resources & Mining" (code for the oil business).

....

Texas is doing well for a lot of reasons. Oil is about 7% of it. Being close to Mexico, having good natural seaports and warm weather and so forth, is probably another small portion of it. Really, though, the vast majority of Texas' current success is attributable to its overall tax, business, and regulatory climate.

That was 2008. Three years later, the nation's economy is much worse, Texas' is also worse, but Texas has hung in there, relatively speaking.

Three years later, oil remains a part of the Texas economy, but there is not enough clear evidence for such a clear throwaway line about how Texas is lucky that oil prices are high. Incidentally, oil and natural gas prices are much lower than they were in 2008:

oilpricessince2008.gif

Texas' job growth has been across many sectors:

jobgrowthacrossallsectors.gif

It's an intellectually lazy cop out (and just plain incorrect) to say that Texas is only doing relatively well, economically, because of oil.

-------------------------------------

Previous Trivia Tidbit: Congress Destroys Wealth.

rss-icon.gif twitter-icon.gif facebook-icon.gif

Posted by Will Franklin · 9 February 2011 11:50 AM · Comments (1)

Trivia Tidbit of the Day: Part 921 -- Congressional Wealth Effect In 2010.

When Congress Is In Session, Stocks Perform Worse-

Interesting data:

congressionalwealtheffect2010.gif
2010 was a good year for the Congressional Effect Fund. CEFFX finished the year up 15.20% as compared to the S&P 500, which was up 15.06%. Equally important, because we were only in the market when Congress was out of session, this performance was achieved by taking much less risk than the broad market. Using monthly volatility to calculate an annual risk, last year CEFFX had a standard deviation of 10.70% as compared to 19.30% for the S&P 500 Index.

This same comparison shows in our longer term numbers too. From inception through the end of 2010, CEFFX had an annualized return of 2.88% while the S&P 500 Index had an annualized loss of -1.65%. During this period, CEFFX had an annualized Standard Deviation of 10.51% as compared to the S&P 500 which had and annualized Standard Deviation of 23.02%. In other words, since inception, investors in our fund had less than half the annual risk of the broadly diversified S&P 500 Index.

2010 was not really a banner year for the Congressional Wealth Effect, since the market fared somewhat well, but the longer term picture definitely still indicates that Congress destroys wealth by its very existence. Being in session means stocks will perform worse on those days, generally speaking.

UPDATE:

Old posts on this phenomenon:
"Congress Destroys Wealth"
"Congress Ruins America"
"Congress Ruins America"
"Unchecked Congressional Power Bad For America"


-------------------------------------

Previous Trivia Tidbit: Cuts Aren't Terrible.

rss-icon.gif twitter-icon.gif facebook-icon.gif

Posted by Will Franklin · 8 February 2011 07:48 PM · Comments (0)