Wednesday, October 22, 2008

CACM 51(10)

In the latest CACM, this article caught my eye, since I have in the past played Go competitively, and tried my hand at Go programming. The article reports that the Go-playing program Titan defeated a human expert, Titan having been given a 9-stone advantage. (An earlier press release is here. This report gives more details, and has a link to the game itself in SGF format. Mick's computer go page keeps track of events like this, and gave me the previous link.)

On the one hand, I like to hope that Go engages the human brain in mysterious ways, so that programs will never catch up with human experts. On the other hand, I suspect that Go will go the same way as chess, and computer programs will eventually win. It is reasonable to hope that will need more emphasis on smart algorithms than on raw computing power. But, regarding the prediction in the article, my guess is that programs won't beat human experts by 2020.

Elsewhere in this CACM, is an article my my former colleague at Warwick, Martin Campbell-Kelly, on the history and future of open source software.

Unrelated: this article in the latest Times Higher is about UK academic blogs, but this blog doesn't get a mention :-(

Tuesday, October 14, 2008

why LaTeX is like Latin

I got my graduate student to stop sending me documents in MS Word, and start using LaTeX, but my objection to Word didn't get to the heart of the problem. Of course, mathematical expressions look bad in a Word document, and the typesetting in generally worse in other respects, but it would be wrong to focus solely on appearance.

The rationale for LaTeX is analogous to a common cliché about the use of Latin, which is that Latin supposedly lets one express things very precisely. One thinks of old-fashioned lawyers making that observation to justify their use of Latin tags, although I don't know enough Latin to say whether that is valid. But, in the context of mathematics and science, MS Word provides a kind of license for sloppiness that is denied by LaTeX. Symbols, especially when the have superscripts, primes, and similar embellishments, do not get a consistent appearance in Word. Extra space amy appear before a superscript, or sometimes the font size changes, because the author doesn't have enough control. In LaTeX, any symbol has a corresponding sequence of ASCII characters in the source file, and there is no ambiguity about whether two symbols are the same, or different. Furthermore, features like the "theorem" and "proof" environments force us to state our theorems and proofs in a self-contained way, rather than having discursive paragraphs that segue vaguely into something that could possibly be interpreted as a theorem statement.

LaTeX, like Latin, does not prevent one from writing down a statement that is incorrect. (A widely cited example is "Dulce et decorum est pro patria mori.") But it does help to prevent vagueness.

Monday, October 06, 2008

Time for smugness?

In the movie Mary Poppins, there is a scene where the admiral asks Mr. Banks how things are in the banking industry, and his smug reply goes something like "credit has never been stronger, and the British pound is the envy of the world!"

After a quick perusal of other blogs that I occasionally read, I find little evidence that the academic community thinks anything to the contrary. This reference to the banking crisis is a rather tangential one, this link from Mitzenmacher's blog. (Conclusion: if you have tenure, sit back and enjoy as the rest of the world crumbles.) This post at Luca Aceto's blog briefly refers to financial turmoil, towards the end. (A blog based in Iceland seemed like a good place to search for comments on this topic.) (Added 15.10.08: this post at the Fortnow/Gasarch blog now addresses the topic.)

Articles here and here tell a pessimistic story, focussing on the increase in pay costs faced by UK universities, due to the increase in inflation. But the banking crisis itself is more likely to impact on the Universities Superannuation Scheme.

Like other private-sector pension schemes -- indeed, like its US counterpart TIAA-CREF -- the USS is underpinned by its stock market investments, and recent declines in share prices have opened up a hole in its finances. Plainly, the right thing to do is for USS to raise the contribution level from staff. Tenure or not, this amounts to an effective pay cut, but the only alternatives are a sudden major recovery of the stock market (unlikely), or a decline in academic life expectancy (hopefully also unlikely). (Note: the situation is different for TIAA-CREF, which is a defined-contribution scheme. For that, the value of your pension goes up and down with the stock market, and it's up to the individual to decide how much more they need to contribute to it.)

I recommend Robert Peston's blog for insightful commentary on the economy.

Wednesday, October 01, 2008

WINE 2008 papers

As a member of the WINE 2008 program committee, I hereby exercise my right to advertise on my blog, the list of accepted papers. If I appear to be "late with the news", my excuse is that I was waiting for the list of short papers to appear, not just the list of long papers. For your inconvenience, both lists seem to appear as Excel spreadsheets, doubly annoying if, like me, you previously made the mistake of installing the odious Office 2007 in place of the slightly more benign Office 2003.

Moving from presentation to content, I see an continued emphasis on non-cooperative games, and in particular, congestion games on networks is still going strong. There's a bit of social choice theory, like "Anonymity-Proof Voting Rules" by Vincent Conitzer, and "Frequent manipulability of Elections: The Case of Two Voters" by Shahar Dobzinski and Ariel Procaccia. I do not count "Biased Voting and the Democratic Primary Problem" (by Kearns and Tan) in this set, since there the emphasis is more on information diffusion in networks, than on properties of voting rules. There are one of two papers on cooperative games, but not many. (e.g. "Overlapping Coalition Formation" by Chalkiadakis, Elkind, Markakis and Jennings.)

On a completely different topic, it nice to read about Richard Stallman having a go at "cloud computing". His complaint, that it's a mechanism to suck users into locked, proprietary systems, looks entirely valid. Also from the article: His comments echo those made last week by Larry Ellison, the founder of Oracle, who criticised the rash of cloud computing announcements as "fashion-driven" and "complete gibberish". Indeed. I would personally criticise the phrase "cloud computing" itself; and note that any intended metaphor is misleading.