About
Yet Another Self-Important Blog.
Rohit Singh
Email Me
Blogs
Use as directed or at your own risk.
Links
These are a few of my favourite links.
[Fri, 24 Oct 2003]
Jabberwocky...
Once upon a time there were big computers which took punch cards. These
amazing machines could do calculations amazingly fast. People wrote
programs to do cool stuff with these things- like proving simple theorems. Soon
enough, there were predictions about how computers could do all the hard
tasks. They would play chess. They would be put in really smart robots.
They would also be used to translate documents in Russian into English so
America could keep tab on what the Russians were upto. Of course, it was
just a minor annoyance that the existing programs were still a bit rough
around the edges- a Russian sentence which said "The wine is good but the
food is bad" was translated to "The spirit is willing but the flesh is
weak". What went wrong was that while comps are pretty good at doing
complex math stuff and logical stuff they are not yet good enough at stuff that is
intuitive to us- our common sense, our visual abilities, our ability to
generalize from examples to theories etc. Of course, AI paid for the
bragging. Funding died down in the 80s and early 90s and is only now
coming back when people have scaled down goals and are trying just solve well-defined subproblems well. Microsoft Office's clippy is a somewhat topical,
if not glowing, example of the current piecemeal approach. One of our
profs at Stanford mentioned that the guy at Microsoft Research who did
work on the Bayesian Networks ideas underlying the clippy is horrified at
his version of the Frankenstein Monster.
The Loebner prize has an interesting history. Early on, somebody (I forget the name) created a program that would chat back to you. He invited his secretary to try it. 5 minutes later the secretary, slightly uncomfortable, looked at him. Could he please leave the room- she wanted some privacy! Ah well, the program was actually quite cute- it'd throw parts of your sentence back at you, along with some suitably vague conversation keep-alive-ers. ("H: I am sad? C: Why are you sad?") But it was really stupid otherwise. In the 80s/90s a guy called Hugh Loebner set up a Loebner prize: invent a chat-bot that would fool humans. The prize was worth $100K. Of course, until that ambitious goal is achieved, a prize of $2K will go to the best chat-bot at the annual competition. The AI community, predictably, was enraged at this trivialization of their research: not only is the goal of the Loebner Prize totally unachievable (atleast for now), the work involved in it is not "useful" research- it is just a bunch of hacks. Marvin Minsky, one of the greats of Old AI, set up a prize of $100 for the first person to get Hugh Loebner to revoke the prize. In a classic move, Hugh Loebner advertised this as Minsky co-sponsoring the Loebner prize: the final winner will also "get" Loebner to stop the competition and win $100,100 (the $100 coming from MM) . Academic fights are so much fun.
Anyway, here's the winner of this year's Loebner prize: Jabberwock. See if you can fool it. (Hint: try using some garbled words). On a related note, but one thats a lot more pleasant to real CS researchers, see if you make out real photos from computer-generated images in this set.
Only in America
A gets a restraining order on B. B sues A for distress. As an added irony, notice that the American papers didn't even find this newsworthy.
As Rummy said, so be it
Spain really wants to
move towards a "new Europe" ;-) They are tired of letting big brothers
Germany and, especially, France do the talking. This chest-thumping
probably also has to do with some of the EU politics going on- in the
current EU setup Spain got 26 votes against France and Germany's 28 (based
on population, I'd guess). Under the new EU setup (more countries are
going to join), these votes will *actually* matter and France/Germany
don't like Spain having so many votes- its economy is nowhere close to
theirs. Spain, of course, likes things just the way they are...
BRICs in 2050
A few days ago somebody mentioned on the Stanford desi mailing list that
a new Goldman Sachs study predicted that India would be the 4th largest
world economy by 2050. Predictably, there were a few sniggers- not because
people didn't believe him but because only a self-important I-banker would
attempt to do such a long-range study. Could anybody predict in 1946 that
USSR won't exist in another 50 years!!
Sometime later, I ended up at the Goldman Sachs website and got the real stuff. To their credit, they had tried the same methods at the 1960 time-point and got a decent enough answer, in terms of their projections matching the current scene. So here's the basic story: in 2050 China, USA, and India will be the 3 (by far) largest economies followed by Russia and Brazil (BRIC: Brazil, Russia, India, China). Each of these economies will overtake the individual G6 countries and the rest of the G6 (G6= G7-USA), put together, at various times. China is already set to overtake Britain and Germany in the next few years and India will be following closely (about 7-10 years behind). Of course, this doesn't mean that the per-capita incomes will be anywhere as close. India's PCI will be $17K in 2050, 35 times today's but still only the same as Portugal's PCI *today*. India also has a younger demographic profile and it will be the only country whose GDP will be growing significantly in 2050. US will still have, by far, the largest PCI.
Now for my two cents :D. Here's my basic assumption, something that others have also mentioned: the last 20 years have changed the rules of the game. The fact that information can be shifted so cheaply and painlessly and that collaboration across vast distances is a snap implies that the barriers of entry, economically speaking, for a country have decreased. Bad phone system ? No problem, get cellphones. Bad manufacturing ? Go for services. Of course, roads and power are still inconvenient imperfections- thats why India will take more time to build up momentum than, say, China. But wait a sec!! What about the current set of rich countries ? Won't they be getting richer as well?
Its almost a cliche that "we'll out-compete them by moving up the value chain". Really ? How high can you go ? Any population can have only so many smart people. What about others who can't climb the value chain to that height ? Where will they go ? Actually, Businessweek's Executive Editor wrote an editorial on precisely this, with a lot of hand-wringing. I can't find the link :-( Anyway, the point I am driving at is: a business' location will become a function of its ability to find the best employees, not the customers (this is a stolen idea, from a book I am currently reading). This directly implies that a country's progress will become, more and more, a function of its population size, provided that it manages to get some basic things right. This importance of population has been true even before- the reason why America is so huge economically is not just that it has enormous resources but also that are a sufficiently populous country (cf. Canada). Of course, America is an exception- their economic policy is also better than most other countries. But the idea still holds.
Another corollary of increased international trade and cheap communications is that that countries don't need to build their economy totally from the ground up- each can start off by competing on its strengths (services, agriculture, tourism etc.). Thus, its easier to play catch-up. And, in per-capita basis, there's a *lot of room* for the BRICs to catch-up. At the same time, the large populations of the BRICs will mean that modest PCI increases will translate to huge GDP increases.
Of course, this is dependent of a huge lot of ifs. Until now, America is the only country to have gotten most things right, consistently, for the past century. They are going to be at the top for quite some time to come. Actually, I'd take America over China as the top-dog anyway.
This sort of ties in with the rather the-sun-is-shining stuff that has appeared in Indian papers recently (1, 2)