Is 2009 too late to start a new blog, something with more than 140 characters per post?  I like to consider myself an early adopter of technology, but somehow it’s taken me until 2009 to get going.

I’ve been working with NLP and machine learning lately, and I’ve had a flurry of thoughts on how to use these technologies to build a smarter web, a web that seems to think for itself.  Building intelligent systems is an incremental processes (whether done by human hand, or the hand of evolution). Between systems with no intelligence at all and those that think, there are many possibilities with intermediate complexity and intermediate intelligence. These systems do not “understand” or self-reflect in any interesting sense, but they are based on many of the same building blocks, and principles as human cognition: statistical learning, distributed architecture, fuzzy representations, adaptive heuristics, developmental assembly, evolutionary design. I’m less interested in far ranging predictions about strong AI than in these intermediate systems that we can put on web right now.

Pluribo was an example. There is so much user generated content on the web. I wanted build a system that could scan the thousands of user comments about a product or topic, and generalize or summarize this content much like a human editor would. Pluribo is imperfect and it doesn’t generalize through the same cognitive process as a human, but it does generalize. I can imagine a series of modifications which would make the system more robust and naturalistic over time.

There are many of these systems I want to build, and I can’t realize all of them (not enough time, not enough money, and many of them probably aren’t viable anyway). But I can work on some of them, and I’ll occasionally write about the others here.

Advertisement