[Computational Complexity] ICS I: snapshots and the longer view (guest post)
- (Guest Post by Rahul Santhanam)
Title: ICS I : Snapshots & The Longer View
1. Local Arrangements: Kudos to the organizing committee for going far beyond the call of duty and arranging for the coldest Beijing winter in 40 years. By deterring sightseeing, this ensured healthy attendance at talks and more opportunities for informal interactions among the participants.
2. Los Amigos: The surprising chilliness of the weather outside was balanced by an equally surprising warmth indoors. High and low mixed with each other, strangers did not scruple to say hello. Key gambits from FOCS and STOC such as the unrecognition (looking right through someone you've met many times before) and the Arctic Smile (the frozen mask whose meaning is - "There's nothing I'd like more than not to speak to you") seemed absent. Not that these gambits are signs of hostility - rather, in our community, they seem the result of social awkwardness together with a consciousness of the limited time available at conferences for attending talks, meeting friends and proving theorems. ICS was friendlier in part because it was more relaxed, but also because it assembled a new "social configuration", with fewer established cliques and hierarchies.
3. Law of the Excluded Middle: There were a lot of distinguished attendees at the conference - people who've had seminal ideas and founded entire fields, but also a significant younger crowd of grad students, postdocs and post-postdocs. The "middle level" of people at the post-tenure, pre-professor stage were rather sparsely represented, though. Maybe this will change when the conference becomes more established... I do hope it's not an indication of a philosophical difference between generations about what constitutes valuable research.
4. The Dilettante Has Qualms: Which of us haven't seen (or for that matter, haven't been) a conference dilettante - someone who makes sure to attend their own talk and maybe two or three others chosen at random by flipping through the conference proceedings, and spends the rest of their time productively in shopping and sightseeing? It is not possible to completely eliminate the dilettante, but it is possible to discourage him, to give him qualms. The speakers at ICS did a fine job of this by motivating the talks so well - no longer was it acceptable to stay away by claiming you knew nothing about topic X and hence were likely to get anything from a talk on the topic. The fact that a talk was on a completely different topic was almost an incentive to attend. Of course, you did run the risk of having to give up your prejudices about those strange other subfields of theory - "trendy" or "incestuous" or "esoteric", as the case might be.
5. What is New?: So what was new about ICS as a conference ? The face-mask dance! Edible conference food! Re-imbursed conference costs! True, and true, and true, but the question was more about the basic structuring of the conference with regard to talks and sessions. Early on in the process, it seemed there was an initiative to have several panel discussions to supplement the talks. Eventually, we had just one panel discussion, and the talks were pretty conventional, but this is understandable in the first edition of a conference, when the conference is still trying to find its footing. What I would like in ICS talks in the future is more audience participation. You can't agree or disagree with the proof of the parallel repetition theorem - it's just there. With so-called conceptual talks, on the other hand, the speaker usually needs to make a case, with regard to the validity of a new model or the importance of a new perspective. This is best done in a dialogical framework, as in economics talks, where questions and disagreements are plainly voiced. This does make it harder to fit talks into fixed time slots, so maybe it's worth looking at more flexible scheduling...
6. The Hare and the Tortoise: My favourite talks at the conference were Srikanth Srinivasan's (on work with V. Arvind about a connection between circuit complexity and the remote point problem) and Ariel Gabizon's (on work with Avinatan Hassidim about derandomization of streaming algorithms and communication complexity protocols on product distributions). Both talks were excellent, but while Srikanth's was more of a standard-issue talk that was exceptionally clear, Ariel's was distinguished by a stylistic tic. In the middle of each slide, he would pause almost theatrically for a few seconds, as if to allow the audience to absorb what he had just said. Ironically, while the talk would have been memorable even otherwise because it was well-structured, this novel (to me) feature of the talk will make it unforgettable. I'm so used to theorists viewing time as a constraint and making the most of every second they have available; Ariel's tactic of having time make its presence felt in a positive rather than negative way seems liberating.
I was told just before my second talk that the time slot had been cut down from 30 minutes to 25 minutes, and when I expressed my worries to my session chair Mike Saks about how I'd adjust, he advised me to speak 6/5th as fast, referencing an old story about a Narendra Karmarkar talk. Maybe it's time now to start preparing 15 minute talks for 25 minute slots and to speak 5/3rd as slowly. I remember reading once during Obama's election campaign that the power of his speaking style owed a lot to the measured way he spoke - speaking slowly is a sign of confidence, and what is said becomes, ahem, more momentous. Come to think of it, we already do have a speaker in our community - Ran Raz (perhaps not coincidentally, Ariel's PhD advisor) - whose talks illustrate perfectly the virtues of simplicity, clarity and not being rushed.
Posted By GASARCH to Computational Complexity at 1/13/2010 09:05:00 AM