Sunday, November 15, 2009

TopCoder: lying again.

background: TopCoder is a programing spec work business (spec work is called "crowdsourcing" nowadays). They also run some regular programming competitions (which are not work for hire), sometimes with okay problems, sometimes with so-so problems, sometimes with problems that 1/3 of participants can solve exactly. I competed there a little once in a while purely for self evaluation purposes - they do somehow have a big community, and there are a few good programmers on top to compete with.

Anyways, where was I... yes, TopCoder lying in their press releases.
It's interesting how a company can't learn a lesson that lying in public releases is not always a great idea. A while ago, they had hired some girl in china - she may have been a good choice for the job - I've no information about this - and then lied a shitton about her qualifications and achievements [see original TC's press release which was then echoed by girl's university] resulting in massive PR success followed by even more massive PR fail in the china, totally ruining girl's reputation. The lying, for a public release, was not very outstanding - just massive exaggerations, pretty standard for small company's public release, a small company has to look big, but it did ruin the girl's reputation 'cause of cultural misunderstanding, its not everywhere customary for a company to exaggerate how great their new hire is. On darker side, I bet they got her to sign their "affidavit" beforehand which explicitly forbids you from suing TopCoder for damages arising from this sort of misrepresentation of you. [you need to sign this at notary if you participate in competition and get a prize; that's quite serious. I won a prize at TopCoder once and asked for legal advice on their affidavit, a friend told me of that girl's story, which I remembered 'cause its really scary how individual could get chewed up by gears of commerce and spit out]

Recently, there had been a "NASA-TopCoder" contest with '25 000 $ in prizes'. It seemed a little strange.

The NASA-TopCoder Challenge will be the first time the TopCoder community of more than 220,000 software enthusiasts is utilized by the world's leading aerospace organization. Long-term human space missions such as those being planned for Mars, will require higher levels of pre-planning and more analysis of available data than ever before. Biometric modeling and simulation programs are algorithmically-intensive as flight surgeons explore and evaluate every possible medical scenario that might occur on long-term missions. In this experiment, competitors will develop algorithms to help NASA's flight surgeons make decisions involved with optimizing the contents of the medical supplies kit that may one day be carried onboard long-term space missions. The submissions will be compared with the results of an existing computer model that has simulated the expected medical occurrences and outcomes for various mission scenarios.

Under closer examination (I registered for the contest because I was rather curious and because invitation email didn't quite made it clear who funded the experiment), it turned out that it indeed was a business research experiment (25000$ from research grant from some business university were used to run 24 tiny contests in parallel for some sort of business research). Needless to say, there were no NASA representatives on contest forum answering the questions about problem or asking questions about solutions [correct me if any did show up since I lost the interest]. Nothing of this sort. Typical programming competition, with a typical competition problem that has only superficial resemblance to real requirements for real software. Very simple model - much simpler than your 'model' when you visit pharmacy and decide what to buy. In real life if you get a splinter under your skin, you will need tweezers to remove it. Then you can use hydrogen peroxide or you can use iodine, or other antiseptic, and if you don't treat the cut with antiseptic you might need to use topical antibiotic later to treat inflammation. That is not simulated in contest - the supplies are not ever interchangeable and medical conditions are not dependent on prior conditions and treatment (worse than that, them are totally statistically independent from prior conditions). It's absurd to think that contents of medical kit for a Mars mission would be based on such simplistic assumptions, so much more simplistic than the ones you'd make when you visit pharmacy. Yet participants would believe it because it's happier to believe you contribute something to space exploration.

Furthermore, interestingly enough, in the "community of more than 220,000 software enthusiasts", only about 1700 registered and only 400 participated in the contest.

As NASA source indicates, the truth boils down to this:
The competition originated when professor Karim R. Lakhani of Harvard Business School and professor Kevin Boudreau of London Business School invited NASA to provide a compelling technical challenge to monitor and analyze the results from an open innovation management perspective. Their research project is funded by grants from the London Business School M-Lab and the Harvard Business School.

, and naturally "topcoder asks for and gets a simple contest-style problem from NASA for use in their business experiment" is a whole bit less impressive than "NASA employs topcoder to solve something for human spaceflight".

2 comments:

  1. From the TopCoder Forums on 6/24/2010

    Hi TopCoder community! My name is Bara Reyna. I am in charge of all of the medical hardware on the International Space Station and also responsible for coordinating the Med Kit Optimization algorithm competition with TopCoder and Harvard.

    I first wanted to thank everybody who participated in the NASA/Harvard/TopCoder competition. It was extremely exciting for me to participate in the competition because this is a VERY, VERY new and different thing for NASA. We typically keep to ourselves when we work on projects, but for this, we were able to literally work with folks from around the WORLD! How cool is that!

    Next, I wanted to apologize for the lengthy gap between the end of the competition and now. I had to get all sorts of approvals, but now that the approvals have been obtained, I will be able to post with more regularity. My plan is to monitor the forums for the next few weeks and respond accordingly every other day or so.

    I'd also like to tell folks how we have used the code developed during the competition. After the competition was done, we went through the top 5 highest scoring code submissions. We looked at the actual code and we looked at the documentation. To put it very bluntly, we were blown away with the results. The amount of useful code developed in such a short amount of time really made us reconsider some of the ways that we write software. We weren’t able to directly import the code into our software because our model is using SAS, but we converted elements from numerous submissions to develop an altogether new SAS algorithm. This algorithm works like a dream.

    We are currently using the algorithm to redesign the contents of the medical kits that are rotating around the earth as we speak! We are finalizing the design of the kits and contents this summer and hope to launch them on a Russian launch vehicle early next year.

    Finally, I wanted to let you know about some future plans. This competition was a pilot project to see if an algorithim competition is a valid way for NASA to do business. The result was a resounding, Yes, We can! We are currently talking to NASA headquarters to see if we can set up another round of competitions, but nothing has been finalized and there is no real time table. I will, however, keep the TopCoder community up to speed on our progress.


    Thanks again for your contributions.

    Bara

    ReplyDelete
  2. my respect for nasa just dropped a few points. A model that is insufficient to describe my purchasing decisions on pharmacy, useful for NASA med kits. I suspect magical thinking that choosing a medkit with software is superior to letting an expert choose it.

    ReplyDelete