Today is the official opening day of GWAP: Games with a Purpose. This is one of two research projects I have been working on for the past few months, though my involvement with GWAP so far has only been in the form of attending meetings, minor testing, and offering my sage gaming advice (and by sage, I mean the herb). GWAP is the next phase in Luis von Ahn‘s human computation project. If you visit and play some games, not only will you be rewarded with a good time, but you’ll be helping science! Science needs you. To play games. Now.
The Idea
Artificial intelligence has come a long way, but humans are still far better at computers at simple, everyday tasks. We can quickly pick out the key points in a photo, we know what words mean and how they are related, we can identify various elements in a piece of music, etc. All of these things are still very difficult for computers. So why not funnel some of the gazillion hours we waste on solitaire into something useful? Luis has already launched a couple websites that let people play games while solving these problems. Perhaps you’ve noticed the link to Google Image Labeler on Google Image Search? That idea came from his ESP game (which is now on GWAP).
The Motivation
What researchers need to help them develop better algorithms for computers to do these tasks is data. The more data the better. Statistical machine translation has improved quite a bit over the past few years, in large part due to an increased amount of data. This is the reason why languages that are spoken by few people (even those spoken by as few as several million) still don’t have machine translation tools: there is just not enough data. More data means more food for these algorithms which means better results. And if results don’t improve, then we have learned something else.
The Solution
Multiple billions of hours are spent each year on computer games. If even a small fraction of that time were spent performing some task that computers aren’t yet able to do, we could increase the size of the data sets available to researchers enormously. Luis puts this all a lot better than I can, and fortunately, you can watch him on YouTube (below).
So, check it out already.
I played the GWAP games today and also think these are fascinating. I’m also struck with von Ahn’s work and earlier dove tail parallel’s with two heterodox and also ill fated AI researchers, Push Singh and Chris Mckinstry. There are parallels, perhaps simply zeitgeist, with both Push’s earlier MIT work and Mckinstry’s Mindpixel project and other musings. I just noticed that Mindpixel, Cyc and Singh are all quoted in an I believe earlier von Ahn paper “Verbosity: A Game for Collecting Common-Sense Facts”. Von Ahn is definitely onto something here – he won the MacArthur and this does seem to be a particularly productive line to explore – I suppose this whole constellation (Lenat, Singh, von Ahn, Mckinstry) has further room for exploration. There’s an interesting sci fi novel I’ve been meaning to read about this gameplay regarding a little boy (the name currently escapes me) and of course, Herman Hesse’s much earlier “magister ludi”". The Flash programming like environment for GWAP is also very nice.
By the way, there’s another recent paper by Deb Roy (MIT) that dovetails with some of the Cyc/Singh/Ahn constellation “The Restaurant Game” – Journal of Game development. I have a feelign that there’s going to be a lot more work done in this area. Someone should quickly produce a ‘commercial framework’ for collaborative game harnessing sort of a human SETI game multiprocessor framework so this can be generally repurposed to any envrionment (sort of like XML). Bob Mottram makes a good point that could be used – (i.e. harnesing boolean/opinions)
Hi Ray, thanks for the comments. I’ve thought about the similarity with Singh/McKinstry and Luis before as well. They were both fascinating people (and Luis is as well). Verbosity is really just the tip of the iceberg on the possibilities in this area, and I’m looking forward to future games that carry the torch further. Luis talks about the fact that CAPTCHAs are a sort of arms race against spammers. Spammers advance AI by defeating CAPTCHAs while the existence of CAPTCHAs encourage advancements in AI. The same applies to these games. As the games generate more data and machine learning algorithms improve, the games will need to evolve to produce more useful data.
Yes, I agree. I read von Ahn’s dissertation and a couple papers based out of this and he’s definitely tapping into a more innovative current with the ‘human computation’/game harnessing methodologies and I’m looking forward to how this develops. If you notice any more current Media Lab/Carnegie Mellon Computer Science type dissertations advancing these types of lines and worth looking at further , feel free to shoot me the email. I noticed on von Ahn’s vita he applied for a patent with regards to a game and ‘moving images’ on his vita – I’m looking forward to seeing this eventually realized online. While our metadata and information retrieval methodologies for ‘still images’ are rudimentary, ‘moving images’ and video ‘metadata’ application is at a much more fundamental stage – btw, very interesting unsolved ‘comp sci’ challenge. Someone should do this as a game experiment and do the ‘Bladerunner’ or ‘Godfather’ game and see what the metadata produces for taxonomy schemas. Onwards and upwards!
By the way, the name of the novel I mentioned above was Orson Scott Card’s “Enders Game”. I’m currently glancing through a collection of essays – mid twentieth century mathematician, Stanislaus Ulam (i.e. Monte Carlo Methods). He develops early uses for computers with his friend John von Neumann. Interestingly, Von Neumann writes in a letter/algorithm for the original statistical recipe for neutron diffusion calculation:
It would require a not inconsiderable number of computers for several days per criticality problem, but it may be possible, and it may perhaps deserve consideration until and unless the ENIAC becomes available.
What is interesting here for me is that Neumann uses “computer” and later ‘computer team’ early I believe here to refer to the ‘human being who will be carying oot the the digital work until the ENIAC becomes available (May 1947). von Ahn I believe is somehow capitalizing and looking deeper into this irony from a different angle. Ulam notes one of their earlier mentors remarks (Stefan Banach) “Good mathematicians see analogies between theorems or theories; the very best ones see analogies between analogies” – on Ahn is already synthesizing and making subversive analogies between “Hard Classic Computer Science Problems” (image retrieval, AI, aesthetics) with the unrealized potential of networkeed “Human Processing”/Games – the trick now is see the analogies between analogies.
Yeah I’ve heard a couple ideas regarding tagging/annotating video, but I can’t remember who told me and under what sort of confidentiality, so best to keep my mouth shut until I can find references. Very cool stuff, in any case.
Well, its very interesting how knowledge transfer and synchronicity works. I begun with the von Ahn Utube Google Video around six months ago, then reading his dissertation about three months ago and now as the university for which I’m in charge of the libraries information system had a ‘spam’ attack of our forms and so we ‘coincidentally’ ended up implementing von Ahn’s stronger and strangely appropriate ‘recaptcha’ technology on ‘forms’ http://library.uwf.edu/Services/OnlineForms/SuggestionForm/SuggestionForm.php, blog http://librarydigitalservices.uwf.edu/library/ and recent staff task force wiki http://librarydigitalservices.uwf.edu/tf-wiki-1.12/index.php?title=Main_Page to better prevent further attacks. Well, I suppose what begins in ‘visionary’ ends in practicality – research to praxis! I would have never come across the “recaptcha’ quick possibility for implementation (we needed it immediately) if I had not gone wide with his dissertation and gwap site! Interesting how consciousness works and knowledge is transferred! One of my techs here also who is working on our longer ‘recapcha’ form list also made me aware at this time of Amazon’s Mechanical Turk http://www.mturk.com/mturk/welcome and Stephe Fosset Article http://blog.wired.com/defense/2007/09/geeks-spot-foss.html As I’m currently head of systems in an academic library with a digital library/initiatives program http://library.uwf.edu/About/Automation/Automation.htm I find it fascinating that von Ahn is so strategically utilizing reCaptcha for ‘digitizing books’ for Captcha’s to decipher. That’s the ‘recaptcha’ we’ve just implemented but I wonder how long this has been going on and what the next stages here may herald? We’ve currently implemented an online statististical framework to try to understand how our larger group of users (the entire university) are using our information systems (specifically university library catalog) http://librarydigitalservices.uwf.edu/library/?p=146 but I wonder if this vast search data could also be harnessed (perhaps even on a more universal iinter-university or inter university library level) to create a better form of artificial intelligence with this level of search (different say from Google Image Labeler or say Google Search). Perhaps I’m not asking the questions correctly here or the questions need a better interlocutor but I do believe, von Ahn’s game technology/methodologies could somehow effectively be applied here to harness this vast ‘search’ data which really somehow surveys the ‘western universe of knowledge’ in interesting ways to perhaps create better connections between data and/or people.
So interesting too, that the recaptcha site: http://recaptcha.net/learnmore.html lists WordPress, Mediawiki and Php as the most standard code additions These are the ones we had implemented previously but I wonder why this language and applications were chosen to be highlighted. Zeitgeist.
I’m really intrigued by all the GWAP games. That’s why I made my own game for my Master’s Thesis.
It’s in the same style as Peekaboom. However, the selections are more precise and combined with semantics.
Have a look at it: http://www.name-it-game.com
We’ve recently released a collaborative game similar to the von Ahn work that aims to collect judgements on anaphoric coreference. It’s not as glamorously presented or as inherently interesting a task as tagging images or music, however the results could be just as useful for text summarisation and search engines. We’ve done a lot of work to collect robust judgements in the most efficient way possible.
You can play the game and find out more information at http://www.phrasedetectives.org