Thursday, February 24, 2011

What I Tweeted

Various thoughts not restricted to 140 characters.

Congrats to Mihalis Yannakakis for his election into the National Academy of Engineering as well as new Sloan Fellows Julia Chuzhoy, Rafael Pass, Chris Peikert, Mark Braverman and Anup Rao. TTI-Chicago gets no respect with TTI-C Prof Chuzhoy listed at University of Chicago but it should be fixed soon.

I called the Watson-Jeopardy match in 2009. The IBM segments during the show were great selling points for computer science. John Markoff wrote a nice wrap-up article for the Times but note the correction
An article last Thursday about the I.B.M. computer Watson misidentified the academic field vindicated by Watson’s besting of two human opponents on “Jeopardy!” It is artificial intelligence — not computer science, a broader field that includes artificial intelligence.
I skipped the various university viewing parties in favor of watching the shows with my daughter Molly. During current events in history class she went on a rant why everyone should think the Watson victory was "so cool". That's my girl.

Wolfram Alpha knows Computational Complexity. Pretty impressive but doesn't know everything. Jeff tweeted
When I ask "Is multiplication in AC0?" it tells me about the Ace of Clubs. Fail.
So we need Watson and Wolfram Alpha combined to make sense of our field.

Next week I'm off to Porto, home of Complexity 2012, for some Port Wine, a Francesinha and the thesis defense of co-author AndrĂ© Souto.

6 comments:

  1. I've been pondering for some time.
    No doubt, IBM Watson has done something extraordinary. However, if we look at this as computer scientists, is it really that special? Allow me to explain myself:

    Was Watson's strength its great computing power? If so, then it is a bit dissapointing. I hope that the true power is some new algorithm or the clever use and modification of existing techniques. However, if the latter is true, I wish that this knowledge is being released to the scientific community.

    Are there any papers out there for the work on Watson has done? I know that there was a whole team for Watson algorithms, so I hope that is the case.

    ReplyDelete
  2. At least it knows the easier direction: http://www.wolframalpha.com/input/?i=IP+in+PSPACE

    ReplyDelete
  3. chazisop, Here's a writeup on how Watson works: http://www.stanford.edu/class/cs124/AIMagzine-DeepQA.pdf

    To some extent, yes, it was a triumph of computing power. It could only compete with humans while running on a large number of computers. The algorithms themselves look to be clever, but not overwhelmingly so. Really, I'm most impressed by how well it used a kitchen sink of decent algorithms to create something much more intelligent then its parts. I'm reminded of the Netflix competition, where the best algorithms were the ones that could intelligently direct lots of dumber algorithms.

    ReplyDelete
  4. TTI people are thanking me for getting Chuzhoy's affiliation corrected. I am less sure about the cause and effect. Wolfram Alpha still doesn't know that PSPACE is in IP.

    ReplyDelete
  5. Is this correct that the list of Sloan's fellows in theory are all selected by Eva Tardos and this is essentially Eva's list of the best young researchers in this year. According to sloan
    http://www.sloan.org/fellowships/page/9
    there are only 3 people in the selection committee of CS and only Eva in TCS. Though selected people are very good, but usually for other awards like Goedel, Knuth, NSF CAREER, Best paper awards in conferences, etc there is a committee of a few people (usually 3-5) and there is not only one person who decides such important awards. Any light on this issue?

    ReplyDelete
  6. It would be fun to see how Watson handles a press conference or an interview with Barbara Walters.

    ReplyDelete