Next Sunday October 12th, six different programs will attempt to pass the computer intelligence Turing test (according to the guardian.co.uk).  This also marks a serious attempt to decide the $20,000 Long Bet between Ray Kurzweil and Mitch Kapor, as well as win the $100,000 Loebner Prize in Artificial Intelligence.  In the tests, a human questioner will get to question both a computer and human through a chatroom like text interface.  If the questioner guesses wrong or can’t decide which is human, the program will have passed the test.  Judging from the conversations below however I don’t think the machines have much of a chance this time around… (click though to the article and scroll down to find out which is human and which is machine)

Can you tell human from machine?

One of these conversations is between Professor Kevin Warwick (KW) and a human; the other between Warwick and a computer. Can you tell which is which?

Conversation one:

Subject: Hello. My name is Nick and I am a human.

KW: Are you happy being a human?

Subject: Yes, I am. Are you? Good. Then we are both happy.

KW: Yes, I am. Are you happy with your life at present?

Subject: Parts of it. I hate my job, but most people do, right?

KW: I’m OK with mine. What is your job?

Subject: I work as an ‘online internet advertising monitor’, which is fancy language for electronic filing. What do you do?

KW: I interrogate humans and machines.

Subject: Which ones do you prefer, humans or machines?

KW: Which do you prefer?

Subject: Hmm. Depends on for what purpose you mean.

KW: To go to a restaurant, for example?

Subject: Then I would much prefer going with a human.

Conversation two:

Subject: Hello, my name is Mike and I am a man.

KW: Are you happy being a human?

Subject: Judge, I’m a guy.

KW: Does that worry you?

Subject: Don’t worry, we’ll work everything through.

KW: Do you have worries then?

Subject: Let’s move on from that basis.

KW: Why?

Subject: Why? I like this subject.

KW: What is the subject, do you think?

Subject: What is the subject I wonder if I think? I know that it’s good to thonk about. Think deeplt about mysterious subjects.

KW: What mysterious subjects?

Subject: Think deeply about mysterious dubjects.

KW: What’s a dubject?

I can imagine the next line, “Dubject… Let me tell you about my dubject!” At which point the skin job stands up from the Voight-Kampff test and blows poor KW away…

Photo: John Cage Organ Foundation, Halberstadt, Germany
via ABC News Online

Fair warning for long-term music lovers: the world’s slowest concert, a 639-year organ piece by American avant-garde composer John Cage (01912-01992), will next change notes in just over a month’s time, on 5 November 02008.

St Burchardi church, in the eastern German city of Halberstadt, has played host to the performance since 5 September 02001 (the late composer’s 89th birthday), when it kicked off with 17 months of silence. Cage originally wrote ASLSP (As SLow aS Possible) in 01985. Its maiden performance by organist Gerd Zacher lasted 29 minutes, but Cage didn’t specify a maximum, so in accordance with the piece’s title, musical scholars and scholarly musicians since decided to stage a multi-century version, approximating the lifespan of an organ.

According to a New York Times report on the 5 May 02006 note change, the odd duration and location for this ambitious project, called Organ²/ASLSP, are due to Halberstadt’s cathedral claiming the first organ with a modern keyboard arrangement, built in 01361, 639 years before 02000, the intended start date for the performance. The sustained notes of the performance are possible thanks to the organ’s customised bellows, and tiny sandbags on strings, rigged to hold each note as long as necessary.

The upcoming change will be the seventh chord in the piece. The last change, on 5 July 02008, attracted over one thousand slow-music fans.

(Above, a home video of the most recent note change, on 5 July 02008.)

Blind writer Ryan Knighton was among those present for the previous occasion, on 5 May 02006 (note changes are always on the 5th of the month). In a short but beautifully written article, he recounts his pilgrimage to the Halberstadt organ, as one of the long, tranquil stretches between increasingly note changes draws to a close:

After a few moments standing here, I begin to crave the next note — any note, any change whatsoever. The chord fills my ears, and a kind of audio-claustrophobia overwhelms me. No sightseers or tour guides around to offer any reprieve. Nobody except Justus, an eleven-year-old local boy who took my hand and guided me the final hundred yards to hear this particular sound before it changed. Lucky he was there — once the new chord begins tomorrow, it won’t change again for more than two years. Got to get to church on time.
…
As I move into the monastery and toward the organ, the familiar clicking of my white cane adds texture to the drone. My own noise feels like blasphemy, chatter during a prayer. Granted, this chord has been playing only four months — a frilly little trill in the scheme of things — but that’s no licence for irreverence. I stop again and listen. Perhaps ten minutes have passed, and I’m becoming aware of the chord’s impurities. The faintest blemishes in tone pop and burn away like sparks. The sound heaves and exhales slightly, like the sigh or groan of a weary traveller.

Sighted folks ramble the world to see Grand Canyons and Eiffel Towers, monuments that dazzle the eyes. Because the last bit of my sight could desert me any day now, I asked myself, what would the equivalent pilgrimage be for me? So a white cane tapped its way from Vancouver to Toronto to Berlin and to Halberstadt to hear a single moment out of centuries of sound. One thing giving way to another — the basis of all drama. It seemed monumental to me.

I’m standing in front of the organ now, and what began as noise has become a familiar hum. As I think of the generations who will take care of a song that assumed they would be there to keep it going, I’m reminded of my debt to Justus and everyone else who guided me here. As this song will be, I’ve been passed along from one person to another, slowly, until I made it. Just in time.

~Ryan Knighton, “Monumental Vibrations“, The Walrus, April 02007.

In case you happen to be unavailable for the upcoming November change, you might pencil in a visit to Halbertstadt for February 02009, or July 02010.

Incidentally, Organ²/ASLSP is the subject of Prediction 282, registered with Long Bets in 02007; that the performance will continue uninterrupted to the halfway mark, in the year 02319. Although a clear majority of folks weighing in on the issue have disagreed (18 doubters vs 8 supporters to date), the proposition is yet to find an official Challenger to turn it into a fully-fledged Long Bet.

(Thanks to William Kramer for the lead.)

Prediction markets continue to proliferate. These communities use money to bet on outcomes in the future. If a prediction comes true, the winners reap the money from the losing betters. The price of a bet, or share, fluctuates over time — and thus can be used as a signal for the community’s opinion. In theory a prediction market taps into the “wisdom of crowds,” but can also be viewed as conventional wisdom. However the results of prediction markets have been proven to be reliable conventional wisdom. (See my previous post on the subject.)
There are two kinds of prediction markets: ones where you bet real money, and ones where you bet funny money. Since betting real money keeps people honest (to reduce their loses), markets with real money are considered a much better indicator of opinion than a mere poll — which has no “penalty” for being less than honest. But real money prediction markets are (stupidly) illegal in the US. So token markets like Long Bets and Bet2Give are devised to innovate around the law.
For instance, Hubdub trades token dollars. You are given $1,000 hubdubs at the start, and $20 each day you log on. You win or loose these token dollars on various predictions. There is a leaderboard which displays the highest ranked traders, showing how much they have gained in the last quarter. One fellow gained $1 million hubdubs, and now has a net worth of $3 million. Hubdub dollars are only good for bragging rights.
One clarification of how the price of a bet works (from Hubdub’s FAQ):

If a prediction has a yes value of 43%, does that mean that 43% of people have voted yes?
No, not really. The forecast is dependent on both the number of people who have selected this outcome and the amount they have risked on it. Very roughly, 43% means that 43% of the money risked by users is riding on that outcome.

I was curious how closely the two formats (real and token money) might match each other so I hunted for a bet that I thought most prediction markets might share: the outcome of the US presidential election. From my brief survey, betting real dollars and token dollars give similar results.  More so, there is a pretty close convergence of price among all the prediction markets:
Roughly, the day after Republican VP candidate Sarah Palin gave her rousing nomination speech, all six different prediction markets price Obama winning at about 60%.

Betfair, based in England, trades real money to make bets. It is the biggest prediction market in the world in terms of numbers of bettors and dollars bet. It’s bread and butter are sports events, including the Olympics, and card games, but it also runs bets on almost anything else including politics.
The day after VP candidate Sarah Palin’s nomination speach, Betfair bookies put the odds for Obama winning at 1.6  and give worse odds for McCain winning at 2.72.
Intrade also bets real money, also mostly on sports, but also on many other wagers. On this same day, Intrade money is on Obama winning at 59%.

On this same day Hubdub market rates on Obama win at 63%.

On this same day the Iowa Electronic Markets, which I’ve written about previously, and is the only prediction market in the US to legally use real dollars, has Obama winning at 59%.

On this same day, Bet2Give also pegs Obama winning at 63 cents or 63%.

Bet2Give is run on Newsfuture software and is sort of a non-profit demo for Newsfutures, which sells software for customized enterprise-strength prediction markets. They promise that a company can “harness the wisdom of your crowds.” In Bet2Give you bet with real dollars but your winnings are given to charities, so technically you are not gambling.
Newsfutures itself runs a prediction market using token dollars. On this same day it shows a 60% chance of an Obama win.

PPX is another token market. Run by Popular Science magazine, it is their Prediction Exchange. It does not do political predictions, so there’s no chart or price for a new US president. Instead it focuses on tech and commercial predictions. Such as: Will Netflix top 10 million subscribers by end of 2008? (You need to register to see the wagers).
My conclusion is that token money prediction markets carry the same validity as real money prediction markets, and that they are fairly consistent across markets. In that sense they are probably reliable indicators of what people believe at this moment (not be confused with reliable predictions).

On Wednesday morning September 10th, the very excited and optimistic scientists turn on the Large Hadron Collider outside Geneva…

We have a Long Bet that states “Large Hadron Collider will destroy Earth.” And you can watch the video above of what that looks like.

The “First Beam” will occur at 9:30am at CERN, which I believe will be about 1:30am here in San Francisco as we clean up after the Anathem Event. We will be sure to open a bottle of champagne and have a toast.  The world has been a lot of fun so far, shame it has to end :)

In a rare, if not completely unique, look back at a few of it’s own failed predictions, Wired published the five largest predictive blunders of its 15 year history (see A Look Back).  When we started Long Bets in 02002, it was precisely because pundits and press did not take this type of responsibility.  Kudos Wired!  Now on to the embarrassment…

Stuff that Wired predicted would die (but didnt):

• Commercial Web publishing (April 1996) Online news sites everywhere respectfully disagree.
• Web browsers (March 1997) Push media was about to supersede browsers. Or not. If we could push this claim from the archives, we would.
• Online song swapping (December 2002) Kazaa? LimeWire? BitTorrent? D’oh!
• Futurism (December 2003) Predicting the death of predictions? Niiiice.
• Brands (november 2004) Would someone please tell the tweens — and Abercrombie & Fitch, purveyors of the finest softcore billboards?

One would hope that they are also bragging about the stuff they got right?  If so please send the pointer.

Warren Buffett recently bet an ambitious hedge fund operator $1 million that they won’t beat the returns of S&P 500 after their extremely hefty fees are accounted for. Buffett claims investors will do as well with a no-load index fund over the ten years of the bet. He has long been critical of the performance claims of hedge funds, and his bet is intended to put his money where his mouth is.

Buffett’s million dollar bet was made on Long Bets, the accountability mechanism operated by Long Now Foundation. The intention of Long Bets is to encourage responsibility in prediction-making (by keeping a public roster of predictions), to encourage long-term thinking (by offering a opportunity to shape a long-term bet), and to sharpen the logic of forecasting (by recording the logic of predictions and bets.)

In order to make a Long Bet, bettors need to lay out their reasoning. It’s worth reading the two sides’ very short arguments about investing because the two extremes of investment advice are contrasted in them. Buffett, as usual, is stunningly clear in his argument, which ends:

A number of smart people are involved in running hedge funds. But to a great extent their efforts are self-neutralizing, and their IQ will not overcome the costs they impose on investors. Investors, on average and over time, will do better with a low-cost index fund than with a group of funds of funds.

Buffett’s Big Bet is by far the largest bet on Long Bets. The previous largest Long Bet was one for $20,000 between Mitch Kapor and Ray Kurzweil. The two prominent thinkers were betting whether an AI would pass the Turing Test by 2029. Ray was certain an AI would pass muster by then and Kapor was sure it would not get close. (Incidentally, Kapor told me recently he’s willing to double, triple, or quadruple the bet with Ray, or anyone else betting on an AI by 2029.)

The way Long Bets work is complicated. To avoid laws against wagering, the money goes to charities and not to the bettors. Long Bets takes a portion of the growth in assets being held as its own overhead to adjudicate bets in the future.

In the case of the Buffett Big Bet, the arrangement was an immediate contribution to Long Now from both sides of the bet. The details of what is being wagered and how the results are being decided is complex. They are described in great clarity by Carol Loomis a friend of Buffett and senior editor at Fortune. She debuts the Buffett Long Bet in the June 23, 2008 issue of Fortune. (Even though the bet was made on Long Bets in January 2008, we have not posted it publicly until today, as per the wishes of the bettors, who needed to coordinate the announcement and press attention.)

You don’t need a million dollars to make a Long Bet. The minimum wager is $200, and is open to anyone. No money changes hands until someone takes up your challenge. You can also simply make a public prediction, which does not require anyone to bet against you. Any prediction can become a bet later.

The hope of Long Bets is that these public wagers will prompt people to consider the implications of current developments in the near-distant future — and then to keep their attention on what happens.

Buffett’s bet is an ideal Long Bet. It makes a huge difference to anyone who invests in stocks (as do a large percentage of the US, either directly or indirectly) whether a boring index fund yields as much as fancy private hedge funds. The answer either way would be a huge influential signal. When economist Julian Simon won the famous bet against biologist Paul Ehrlich (Simon betting that the long-term prices of commodity minerals would decrease over ten years; Ehrlich betting they would increase), his win essentially eradicated the argument of resource scarcity from the environmental debate. Environmentalists then shifted their concern to the many other issues needed to foster a healthy environment.

This bet has a similar potential. But as in all great bets, its outcome is uncertain. As Loomis writes:

Buffett himself assesses his chances of winning at only 60%, which he grants is less of an edge than he usually likes to have. Protege figures its own probabilities of winning at a heady 85%. Some people will say, of course, that just by making this bet, Protege has acquired some priceless publicity. But then, Protege clearly wants to win, and it’s up against a man who hasn’t made a lot of losing bets in his life. Seides himself sees one strong ray of light: “Fortunately for us, we’re betting against the S&P’s performance, not Buffett’s.”

The other side of Buffett’s bet is being taken by these guys at Protege.

 The great folks over at Predictify have made a special area in their site for Long Bets.  This is a great place to experiment with predictions and even make short bets that may have long term consequences…  Check it out at:

http://longnow.predictify.com/ 

I’ve Stumbledupon (quite literally) a interesting looking site called Predictify.

Predictify seems to be combining social networking, message board discussion, pay-per-post business models and Wikipedia-style collective wisdom into a harmonious online community of eager questioners and knowledgeable, astute predictors, all united to discuss deterministic questions about the future, share knowledge, and act on the results.

Because mob rule *totally* works on the internet.

With Predictify, registered users can read questions posed by questioners, pose their own questions, predict answers, and repeatedly smack down internet smart-alecks who try to use the site for betting, insider trading, and making off-color predictions about their dorm roommate’s luck with the opposite sex during the upcoming weekend.

Theoretically, predictors also get paid every time their predictions are correct. Questioners can pay a premium based on the number of responses which in turn entitle them to collect value tables for a given question, collect larger data sample sets and keep a running tally of private data if they so desire. The pot is shared out among the predictors who nailed the question most accurately and soonest, according to rank, expertise level, phase of the moon, etc etc.

The site includes live news feeds relevant to the question being discussed. Gimmicks like community points, reputation rankings, and “private prediction environments for you and your friends” trigger that social network addict’s instinct to refresh the page every thirteen seconds.

The truly dedicated Predictify player, if enough time and energy were invested, could find themselves joyously sucked into in an interdisciplinary morass of statistics, arcane behavioral controls, incompatible social networks, philosophical conundra, high level mathematics and a steadfastly erratic human element, which make the site reminiscent of the game Eschaton from David Foster Wallace’s novel Infinite Jest, in that play is aided by the ability to quickly master advanced game theory and do the metaphysical equivalent of pegging small objects with tennis balls with deadly accuracy.

Since this is what we do for fun around here at The Long Now, it sounds like a grand time to me.

Long Bets has arrived at a decision for Long Bet #2 between blogger Dave Winer and Martin Nisenholtz of the NY Times. At stake is US$2000.00 plus half the interest that has accrued over the last 5 years in the Farsight Fund, all of which will go to the charity of the winner’s choice.

In the bet Winer asserts, “In a Google search of five keywords or phrases representing the top five news stories of 02007, weblogs will rank higher than the ‘New York Times‘ Web site”. The premise of this bet is excellent, but unfortunately the arguments were quite vague on how to adjudicate the bet. Long Bets encourages bettors to construct arguments that involve the least amount of interpretation possible. Once this bet came up for adjudication we urged both parties to come to their own decision, but they asked Long Bets to be the final arbiter. We have done our best with the information and resources available to us, but this process should be a good instructor both to future bettors and ourselves…

The major questions that affect the interpretation of this bet:
Q: Which list of “2007 top stories” to use?
A: We chose the Associated Press list, as it was the only one suggested by one of the bettors (Nisenholtz), and it was in effect at the time of the bet origination. We found many others, (some listed in the notes below), that may actually be better indices of what a “top story” is, but we felt that the AP list was our best choice for this bet.

Q: What is a weblog? Does Wikipedia count? What about the NY Times blog or other commercial blogs? Does it include any non-commercial user submitted web site?
A: We decided that a weblog had to be something that would have been recognized as a blog in 02002. This includes ad supported blogs and commercial blogs like those of the NY Times. While the bettors argument in this case discusses why non-commercial content will beat out commercial content, Winer never provides a definition of a weblog. As it turns out, including major news source blogs like those of the NY Times or sources like Wikipedia do not affect the ultimate outcome in the case of this bet, but they certainly could have.

Q: What is the NY Times? Does the International Herald Tribune count (which is owned by the NY Times and its content comes from there)?
A: We determined that it had to be on the nytimes.com web site to count. If the bettor wanted subsidiaries or other associated derivative content to count, they should have specified it in their argument. This did affect the outcome of one of the searches where the IHT.com result came in at 9 and blogs came in at 10. This result would not have affected the ultimate decision however.

Some other notes: The bettors also never defined what the search semantics should be, and or what date the searches should occur on. Both of which affect the data a fair amount. We tried the searches in a number of ways and a number of times since AP released their list of stories in December to arrive at our decision. We disregarded any search results that were dated after 12/31/02007 when calculating search rank.

Here are 02007′s top stories, as voted by AP Journalists with search rankings (lower is better). We also include results of the highest non-commercial/user submitted content and highest ranked commercial content as a reference.

“VIRGINIA TECH KILLINGS” (NYT score 26, blog 10) winner Blogs
Highest user contributed result: Wikipedia 1
Highest commercial news outlet result: USA Today 2

“MORTGAGE CRISIS” (NYT score 2, blog 10) winner NYT
Highest user contributed result: Wikipedia 1
Highest commercial news outlet result: NYT 2

“IRAQ WAR” (NYT score 24, blog 5,) winner Blogs
Highest user contributed result: Wikipedia 1
Highest commercial news outlet result: CNN 3

“OIL PRICES” (NYT score 172, blog 38) winner Blogs
Highest user contributed result: Monga Bay Blog 38
Highest commercial news outlet result: Bloomberg 1

“CHINESE EXPORTS” (NYT score 57, blog 3) winner Blogs
Highest user contributed result: Blogging Stocks 3
Highest commercial news outlet result: China Today 1

• Adding up page rank winners blogs win 4 to 1.
• Adding up page rank winners of user submitted content vs. commercial content, user submitted content wins 3-2.
• If you average page ranks of the NYT (avg rank 56.2) vs. blogs (avg. rank 13.2) Blogs win.
• If you use an average rank of user submitted content (avg. rank 8.8) vs. commercial (avg. rank 1.8) Commercial news outlets win.

The Long Bets decision on this bet is in favor of Winer’s side, weblog page ranks came out ahead of the NY Times. We will be calculating interest and sending a check on to Dave Winer’s charity of choice the World Wide Web Consortium in the next month.

Notes:
Aside from the observation that Wikipedia often ranks very high and was not really considered at the time of this bet in 02002, another interesting note was how well government sites ranked in subjects like oil prices, Chinese exports, and others. The government sites are often listed in the top ten of these types of subjects showing that people are also turning to the government websites for authority.

The other interesting thing to us was how much the bettors own definitions (or lack there of in this case) affected the bet. For instance had the bet been structured around commercial vs non-commercial content, and they had chosen an average ranking system (which actually seems to answer the question being asked more clearly), commercial content would have won by a factor of more than four.

Also of note is that with a slightly different analysis Rogers Cadenhead did come up with the same winning results based on page rank over at his blog Work Bench.

For reference here are some other “Top Stories of 2007″ lists that could have been considered. Testing the first two of these lists yielded results similar to the AP list.

Pew’s Project for Excellence in Journalism’s News Coverage and Interest Indexes.

Time Magazine

About.com

Foreign Policy, top 10 stories missed in 2007

CNN (not ranked – chronological)

MSNBC graph showing top story of the day, for the year (most clicked)

Telegraph UK Top read stories of 2007, by category

Crikey’s Top Ten List

Doctors Without Borders (top *underreported* humanitarian stories):

BBC News (most popular)

Yahoo! News (most emailed)

…by Bill Gates.

Image from the International CES website

Perhaps the central feature of the Long Now’s Long Bets project is accountability — tracking the fortunes of predictive statements and the arguments made in support of them. In a mediasphere with an attention span as short as ours, pundits, CEOs and other “thought leaders” can get away with a great deal of overconfidence and sloppiness in their future-oriented rhetoric.

Which is why it’s encouraging to see, from time to time, reporters digging into the archives in an effort to evaluate a habitual predictor’s track record. (An earlier example blogged here.)

For instance, this Times Online (UK) article published yesterday, 7 January 02008 (“Bill Gates’s hits and misses at seeing future”):

Bill Gates last night described the home of the future where all the power of the home computer will be available in every room on touchscreens embedded in the furniture.

The vision was one of his annual predictions of future trends in technology at the launch of the 2008 Consumer Electronics Show (CES), which opens today in Las Vegas.

But don’t plan on rushing out to buy the hardware any time soon – the co-founder of Microsoft has been wrong as often as he has been right in his forecasts and in the company products he has touted. Some of the moments he might like to forget include:

1995 Mr Gates launches Microsoft Bob, an attempt to out-Apple Apple with a desktop that looked like a house, and software represented as cartoon characters, such as Java the caffeine-crazed dragon. Gates called it a way to make “the computing experience better”, but the product bombed as it needed twice as much memory as the average PC had back then, so didn’t work on most people’s computers.

2001 He predicts – correctly – that by 2010 there will be a PC in 75 per cent of US homes. But he also said that the most popular form of computer would be the tablet, a handheld device which the user writes on with a stylus. Last year tablets accounted for just 1.2 per cent of all PC sales.

2002 Still fascinated with mobile technology, Mr Gates tells CES that Mira, a wireless monitor for the user to carry about, allowing them to access the internet and music files anywhere in the house, meant that “entertainment will never be the same”. It didn’t catch on because it wasn’t compatible with the home version of Windows XP, couldn’t show video, and cost more than $1,000 a go. It was abandoned in 2004.

2002 The Microsoft boss predicts that Smart Personal Objects Technology (Spot), in which pens, watches, coffee-makers and other everyday items are internet-enabled to allow them to relay information such as weather forecasts and sports scores, would be part of the computing revolution. So far it hasn’t been, though it is beginning to show promise for navigation devices.

2004 Mr Gates tells the World Economic Forum that spam will be “solved” by 2006. Er hem.

In his defence, he has also used CES to launch the XBox in 2001, which is Microsoft’s most successful piece of hardware, and correctly predicted that networked mobile technology would become massively more important.

One thing this points up, I suggest, is that in a sense all business decisions are based on predictions, or can themselves be seen as “bets” — something of value staked on an uncertain outcome. And we need not confine our view to business. Every decision we make is a step into the unknown; a wager made, with whatever quality or degree of forethought, that the ground will hold, or (looking further out) that where we eventually end up will have been worth the trip. (more…)