Life & Research Notes

I have a structurally complete version of my thesis now, and I’m just waiting to work on revisions based on feedback from my readers, John, Hiroshi Ishii and ML alum Brent Britton. Though I know I have some more work to straighten it up, and the deadline is coming right up this Friday, it feels really good to have all the parts in place.

The other news is that I resuscitated OPENSTUDIO today, putting it on our newer server, which also involved switching it to apache2.2 and mongrel cluster instead of fastcgi, and setting up Postgres and PHP, plus some Tomcat fiddling. Way too many technologies happening in that project.

Amber and I appear in we-make-money-not-art today in an interview about our work with OPENSTUDIO and in the PLW.

Thanks Sascha!

What a minimal amount of work to switch OPENSTUDIO to public membership! All the code and infrastructure was in place, and it was a matter of just flipping the switch. I made most of the changes yesterday while sitting in the back of John’s class. This morning I updated the site and announced the new policy on the OS blog.

With the original invitation requirement, OPENSTUDIO’s barrier to entry was very high because people could only join the community via invitation from an existing member. Each new member was socially connected to the PLW by an unbroken chain of invitations, each link in the chain offering a property of transitive accountability. Though superficially a small change, open public membership effectively lowers this barrier one very important notch; suddenly we are allowing anonymous, totally unknown users, with no connection to anyone, and no inviter whose reputation is at stake. How will the community develop without the imposed connections? Of course these unconnected users may in turn invite others, potentially leading to more fragmented subcommunities, Chris Anderson’s “tribal eddies.” How will these new dynamics affect the site’s content, the character and quality of the drawings, tags, commissions and transactions?

Of course all these ponderings are meaningless if no one bothers to join. Is there any life left in OPENSTUDIO one year after launch, or are participatory media and virtual microeconomies too mainstream for people to take notice of this funky little art site that’s been humming along in its own little dreamworld? We’ll find out.

This morning Luis and I are having a nice talk about whether an OS-like market should trade in media (drawings, audio, video, animation) at all, or whether we should be thinking about markets for social information: gossip, secrets, tips, howtos, critiques, even opinions. What would you rather buy: a drawing, or an answer?

Last week we finally put up Takashi’s PLWire as the main PLW site. It took a bit of Apache twiddling and coaxing to keep everything else held together, but it seems to be working well now. I’m working on a polymorphic acts_as_permalinkable rails plugin that will make it easy to link directly to the uploaded graphics and videos, and Amber is working on an audiotagging system so that people can call into PLWire and attach messages to these resources. Look for these additions over the course of the next month. And of course if you’re feeling nostalgic for the simplicity of the past you can check out ye old PLW Classic.

Also starting up this week is Burak’s OpenIO site, a sort of social and economic network for hardware devices. Later this month Burak will be presenting this work (among other things) at ISEA in San Jose and then Ars Electronica in Austria.

What sort of processes would benefit from collective human power?

One possibility is image processing: removing wires and unwanted backgrounds from video, cutting out the main characters from a green screen, filters and effects that capture a certain human randomness.

Can we utilize something like the Amazon Mechanical Turk here? I need to talk to John about getting some funds for this.

What sort of image processing tools would make it easy for workers to do this? For each domain (images, transcription/language, trivia/polls, etc) we need a generalized app that can take a remote job description and set up a task-oriented environment.

How can we invert OPENSTUDIO applications to make them for the collective instead of for the individual?

Last night I was talking to Jenn’s friend Sam, a finance guy very knowledgeable about economies and investments. Sam mentioned how certain companies analyze the various “strong buy” “buy” “hold” “sell” ratings that firms apply to stocks. By looking at the history of how a stock has been rated by various firms, these companies can create profiles of the firms that are doing the rating.

When investment advisors rate stocks like this, they are essentially attaching a time-decaying tag to the stock. Similar to OPENSTUDIO’s “bubble up” tagging, these ratings accumulate and begin to paint an aggregate picture of the firm itself. Sam called this sort of profile the first derivative of the actual rating.

One of the key differences here is time. Current tag systems don’t look at time explicitely. At best, such as with del.icio.us, they typically just show a time stamp, and use the time information to detemine current trends and popularity.

In finance time is much more important. Winning and losing only happen with change in the market, which is of course only possible when there is some notion of time. An investment rating is decreasingly relevant as time goes on, with the actual relevance decay rate dependent on the industry, stock, firm, etc. An analyst can renew the rating by reissuing it, or change it. In comparison, tagging is static. Yes, we can change our tags, but the way we change them is not tracked. Aggregate tagging trends can reveal changes in the tagged object, as well as the tagging community. Encouraging change and tracking it is key.

Some ideas that fall out of this:
- tagging stocks
- time-decay tagging

Wikipedia on Prediction Markets