Luis von Ahn launches 'Games with a Purpose'

Luis von Ahn - father of noteworthy projects like the ESP Game (a.k.a. Google's Image Labeler. Radar post) and reCAPTCHA (Radar post) - has launched Games with a Purpose, in short GWAP.

GWAP and the new games - Tag a tune, Squigl, Verbosity, and Matchin, puts von Ahn and his team at Carnegie Mellon by a wide margin at the forefront of tapping into human cycles to do jobs that computers only can solve poorly or not at all.
At FOO Camp last year von Ahn listed the main motivation for his projects as "I hate paying people to do things." Beyond the obvious chuckle such a statement releases, there is a truth hidden to it that Amazon's Mechanical Turk is no doubt feeling and that every HR department knows: people only motivated by money often do a terrible job.

While the ESP Game was licensed to Google as the Image Labeler, von Ahn writes by email that "the data [from GWAP] will be publicly available in bulk to everybody." My colleagues at Polar Rose will be happy to hear that, as am I sure everyone else in this domain will.

Go spend 10 minutes having fun at GWAP and "help the world become a better place" as von Ahn puts it himself.

ps. von Ahn's Tech Talk at Google on Human Computation is a classic worth viewing if you haven't already.

Why Wikia will change search

This morning Jimbo Wales' Wikia launched their search effort, Wikia Search. Wales & co. have been getting a lot of heat for this launch, most notably from Mike Arrington at Techcrunch who calls Wikia Search "an inexcusable waste of time" and "a complete letdown".

I have to be honest that my first reaction to Wikia Search was lukewarm, and I fully support Arrington in his assessment that Wikia Search is no viable replacement for Google, Yahoo!, Ask.com, or any of the other established search players. But Wales takes out the air of Arrington's choler by commenting that "it's a project to *build* a search engine, not a search engine", recalling that

When I launched Wikipedia, I wrote at the top of the first page “Wikipedia, the free encyclopedia”. On that day, anyone reviewing it would have laughed. What’s this? There’s nothing here! This is not an encyclopedia, it is an empty website with some funny editing syntax!

Were Wales goes wrong is that Wikia has released a search engine on a limited index with no opportunity for users to contribute. In terms of Wikipedia, it would be the equivalent of launching an out-of-copyright dictionary but giving users no ability to edit the erroneous articles or adding new. Wikia should not have launched without, and that is a mistake.

Yet Wikia will change search. They may very well be run overend by Google, Yahoo!, or Facebook (who eventually will turn towards search like their fraternal predecessor AOL) in the process of doing so, but they'll change a few rules of the game.

First, values
AOL's search log blunder was notorious, and generally people are slowly starting to question the potential privacy invasion from our online data trails.
A Wikia employee told me today that people were already asking what the most popular search terms were. He said there was no way of finding out as no logs are kept.

Second, they'll open search. Really.
Wikia claims that they'll make their index freely available. If they haven't already, we'll almost certainly see Wikia's index in Amazon's S3 (Amazon is a major investor), making it effortless to create custom search engines using a couple EC2 instances. Think vertical search engines with custom algorithms for anything from gaming to Japanese manga cartoons. Talk about giving Google's 16K employee brute-force machine competition.

Go play with Wikia Search. Then come back here and read the above again. Tell me what you see: a bluff or a ripple of change?

Disclosure: Fellow Radar blogger Artur Bergman is Director of Engineering at Wikia, and I'm posting this just a few hundred meters down the road from their Poznan (Poland) office.

Negroponte: "We're the World Food Program and they're McDonald's"

The title quote is Nicholas Negroponte's in a response to Intel quitting Negroponte's One Laptop Per Child (OLPC) in an interview with David Kirkpatrick of Fortune.

Intel has quit its support of OLPC citing that "OLPC had asked Intel to end our support for non-OLPC platforms, including the Classmate PC, a developing world low-cost laptop running Windows.

Negroponte rebuts that Intel, who was a partner and board member, had not delivered on any of its promises and been perpetually disloyal by in sales meetings claiming that "The (OLPC) XO doesn't work, and you have no idea the mistake you've made." Intel were the ones that formally quit, but they were already on probation and notes Negroponte "If you're in school and you are on probation for very serious misdemeanors you can say you quit, but...".

While Intel's actions as reported are calamitous, Negroponte's comparison of OLPC to the United Nations World Food Program is unfortunate in it's own way. The goals of OLPC are in every way as honorable as those of the World Food Program and it is amazing what Negroponte and his team have achieved, but it is difficult to understand why the OLPC should not operate more in line with the open market.

Last January I found myself in an elevator with Negroponte who was clutching one of the very first prototypes of the XO ("the engineers will kill me if I drop it on the floor"). I immediately asked him "why can't I just buy one? Why a 100.000 unit order minimum?" The obvious answer was that he needed scale, but it was one that I still fundamentally have a hard time accepting.

Mind you OLPC and the low price tag it carries is an ambitious project which has been designed from scratch, but why is it that OLPC should be better at producing and distributing the XO than Michael Dell or Paul Otellini's treacherous - but seemingly efficient - crew?

Back to the question of scale. Negroponte told an audience at Forrester that the OLPC-patented display technology would not be licensed freely in order to secure the display manufacturer amortization of a $2 billion display fab.

While Negroponte is probably right that this was the only way to kickstart things, his next big task is to step back and let the market forces accelerate the distribution to children of laptops built on an open GPL'ed OPLC reference design.

Commoditization has worked great for the PC. Why should the OLPC not make this work in its favor?

"eBay went straight to the nuke"

The title above was makester Phil Torrone's gut comment on the radar backchannel to eBay's decision to cancel all their US-based ad$ on Google in response to Google's 'Freedom Party'. The Freedom party was meant as a friendly but prominent reminder of eBay's unwillingness to let Google Checkout onto their platform.

Fellow Radar-blogger Marc Hedlund quickly pointed to the likely core of Ebay's strong reaction:

I recently read "The PayPal Wars," which describes the growth of PayPal (from their marketing director's POV) from launch to acquisition by eBay. One of the stories was about PayPal crashing eBay Live, when eBay was trying to promote its own payment service. It sounds like Meg Whitman got up at her own conference and saw about 1/4-1/3 of the audience in PayPal shirts -- given away at a party just like the one Google was proposing to throw.
I would bet eBay decided they never wanted to repeat the experience.

Update: Read Dave McClure's comment below for more insight into the PayPal t-shirt revolution party that he took part in putting on.

Google has since canceled the Freedom Party with the excuse that "eBay Live attendees have plenty of activities to keep them busy".

Where's the nimble startup ready to challenge eBay in a powerplay without the risk of losing a three-figure $ million revenue stream, and when will eBay recognize - like Facebook - that they're a platform and not a closed bundle of applications.

reCaptcha - Stop Spam. Read Books.

Carnegie Mellon University professor Luis von Ahn's latest creation reCaptcha is yet another great example of bionic software on steroids.
You'll remember Luis von Ahn as the creator of ESP Game (licensed to Google as their Image Labeler) and derivative works Phetch and Peekaboom, but what is less known is that von Ahn is the person behind captchas, version 1.0.

reCaptcha makes captchas more useful than just preventing spam; by tapping into the reportedly 150.000 hours spent daily typing in captchas, reCaptcha has users proofread book text that OCR could not recognize and which would otherwise have to farmed out to a Mechanical Turk or other distributed proofreader.
Breaking up scanned text into small chunks for distributed processing by humans has been seen in various forms previously, including in banner ads (by inChorus, formerly Mycroft), but reCaptcha's ingeniousness lies in making an otherwise cumbersome task worthwhile every single time.

One can only speculate whether von Ahn have taken inspiration from ingenious spammers who were cracking captchas by asking users to solve captchas inlined from Hotmail and Y! mail in exchange for porn.
He who laughs last, laughs best. Will that be "Stop spam. Read books." or "Read Books. View Porn."?

Angelina Jolie is a Geowanker

According to a story in the Daily Mail Angelina Jolie has taken geotagging to the extreme. The coordinates of her tattoo represent Cambodia, Ethiopia, Namibia, and Vietnam, all locations of birth of her four children.

A quick search on Flickr shows that she's not the only one to sport geo coordinates for love (Jane McGonigal's engagement ring) or personal artifact (Mikel Maron at Lift07).

We'll have to challenge Mikel - geoguru and speaker at Where 2.0 only a week away - to face up the challenge and sport a real tattoo on stage! (For the less brave at heart, temporary solutions here.)

[via Geowanking]

Seeing more with the blind camera

Berlin-based artist Sascha Pohflepp's Buttons is a camera that takes other peoples' pictures:

"It is a camera that will capture a moment at the press of a button. However, unlike a conventional analog or digital camera, this one doesn't have any optical parts [...] The camera memorizes only the time and starts to continuously search on the net for other photos that have been taken in the very same moment."

While the project is already 9 months old and has long since been featured in the online columns of our hip and connected friends at MAKE, Buttons demands all the additional attention we can give it here on the Radar.

This simple project is the prime example of how networked objects allow us to sample off eachothers experience, memory and ultimately senses. The backwardness of the interface is the twisted proof that our own experiences -- even the physical -- increasingly only gain true value when intertwined with the experiences of others.
The more personal sensors we add, the more we'll each be a remix of the collective experience.

Simply brilliant. Makes my heartbeat spike.

[Via classy.dk]

Joost is the iPhone of TV

The Venice Project, Dirk Willem van Gulik (of Apache fame) and Skype wonderboys Janus Friis and Niclas Zenström's peer-to-peer IP TV project, today officially became Joost.

I've been playing around with the Venice player (now Joost player) for the past couple of weeks after finally resolving some connectivity issues. While a bandwidth hog unlikely to work reliably on anything but a really fat pipe, the player works like a charm, quickly loading new videos including a non-intrusive sponsor advert.

The content available right now is pretty limited (a lot of music videos and MTV style content, most likely due to content manager Henrik Werdelin's previous stint with MTV Europe), but it seems only a question of time before more varied content comes available. I know of several production companies in discussion with Werdelin & co. and would be surprised if a tipping point can't be reached pretty quickly, especially in Europe which is dotted by TV production companies heaving to reach audiences beyond the limited national audiences.

My biggest issue with Joost so far has been the generally sub-DivX quality of the video, in my mind ruling out watching more than 15 minutes at a time (I'm sure my eye doctor will agree). Furthermore, there's inconsistency between the image frame resolution of shows, most likely due to the 'crippling' of content by some providers who are still not comfortable with the new distribution model.

Before having tested the player I feared that Niclas, Janus and Dirk might be going wrong this time in that the disruption comes in democratizing the content, not the platform. My argument was that it seemed that they were betting on their business acumen to close deals with the content providers rather than disrupting the actual distribution.

I've pulled back on that conclusion as the genuine disruption lies not in the distribution, but on what comes on top of the TV viewing. Joost is the Firefox or iPhone of TV.
That significant dash of Internet flavor comes in the form of the Mozilla XUL platform, which enables the rapid and simple creation of plugins, already known to a hoard of Firefox extension developers. The Joost developers have already added a few, including a backchannel chat, a Jabber IM client, and show ratings, but the real fun starts when data is moved out of Joost.

Just like most of the really interesting discussion around the iPhone has centered around iPhone widgets, the really interesting discussion around Joost should be the plug-ins.
So how about you: what services did you dream up on the analog tube that were never possible? I vote for a videoscrobbler and matching last.fmtv service, but what about you?

TurkTunes?

Amazon's Developer Connection team is apparently out to take up the competition with music-recognition service Gracenote (formerly known as CDDB). They've released What's That Tune?, an amusing mashup of three Amazon Web Services:

This proof-of-concept records audio from a microphone, saves it to Amazon S3, creates a HIT for [Mechanical Turk] workers to name the tune, then polls Amazon ECS for related albums.

Use What's That Tune? to name the song stuck in your head, to play around with Amazon's excellent pack of web services, or just to be utterly amazed by the Mechanical Turk HITS people are willing to pick up for just a few cents. (See Brady's previous coverage on The Sheep Market)

ps. In other news, Make Magazine's Phil Torrone is asking for the reverse: IBM please open up Big Blue and teach us how computer beat man in the legendary chess world-champion vs machine battle.

Wengo Launches Voice Services Marketplace

France-based Wengo recently launched Wengo Services, a voice services marketplace. Wengo Services allows service providers to set up for-pay VoIP services just like traditional premium rate 1-900 services (known in other countries under other names including 90 and 900 services). Wengo runs a services directory and takes a 30% billing commission, probably not far from what telephone operators charge their 1-900 providers today.

Wengo was originally launched as "an open alternative to Skype" as David Bitton (Wengo CEO), who was also featured at this year's EuroOSCON, likes to put it. The now renamed WengoPhone is a GPL'ed voice-, video-, and IM-client.

The lack of distribution of WengoPhone seems to be the major challenge of Wengo in establishing a worthwhile customer base for it's voice service providers. Seemingly to offset this, Wengo last week launched WengoVisio, an interactive Flash-based widget which allows non-Wengo users to connect to WengoPhone users.
However, while the Wengo Services directory interface hints that there is or will be integration with WengoVisio, for-pay services still require the user to download WengoPhone, a hurdle few users will take upon them for a single call. That and tedious sign-up processes could likely kill off any early success.

Despite current shortcomings Wengo's initiative is still praiseworthy. When eBay bought Skype I speculated that such opportunity was one of the biggest motivations for the acquisition. Yet 15 months later Skype still requires the small-time translator to develop on someone else's development platform, a development platform not even meant for voice pirates but for the telephone operators of lore.

Yet even if Wengo succeeds in making it easy, and Skype facilitates a peer-to-peer charging model, my bets are that the real success will come from the extraordinary Yahoo! Answers. Adding IM and voice would be a plausible next steps and kick Yahoo! in front in the live search battle which is rolling up (Radar post)
If only she were online, I'd consult Wengo's Haitian clairvoyant to know.

Recent Posts

• Simplifying Mobile Testing on December 16, 2006
• Ultimate fabbing experience on November 14, 2006
• Truly Democratic Mapping on October 25, 2006
• The data mining dilemma on October 3, 2006
• Mobile 2.0: Nokia launches WidSets on October 3, 2006
• YouTube and the DMCA on October 2, 2006
• News from the future: Rebooting your car on October 2, 2006
• Life 3.0 - The Hutter Prize on September 29, 2006
• Life 1.1 on September 28, 2006
• Sony Reader finally ships on September 27, 2006