Sunday, March 16, 2014

Compression = intelligence?

Compression = intelligence?  The Hutter Prize is intended to encourage research in Artificial General Intelligence (AGI) by offering (currently) a 50,000€ (about $70,000) prize for solving a specific problem.  The general idea is along the lines of the X prize: by offering a modest prize, one can provoke research that would cost much more than that – along with generating lots of passion, enthusiasm, and competition.  All good.

The challenge with AGI is what one would award a prize for.  So far the achievements of AGI are esoteric, certainly not accessible to anyone other than specialists in the field.  Furthermore, there is legitimate debate about whether even AGI's successes are anything more than trivialities wrapped up in complicated software and fancy words.  Nobody has yet done anything actually useful (or even entertaining) with AGI yet.

So what kind of a goal would be worth a prize? 

The Hutter Prize folks chose lossless data compression of a specific English text file, and I don't know whether to laugh ... or laugh.  Their premise is that solving compression problems is equivalent in some way to general intelligence.  I know a little bit about data compression, and I know some people who are genuinely experts.  I know much less about AGI.  I don't recall ever talking with them about the Hutter Prize, but I'm pretty sure I can predict their reaction: the act of writing a data compression program might require intelligence (it certainly requires persistence!), but the resulting data compression program isn't going to be “intelligent” by any definition of that word!

But if you're into data compression, especially if you're a starving student, maybe that Hutter Prize will spur you into developing the next best data compression algorithm!

No comments:

Post a Comment