Birds and bots: simulation and syntax

John Hawks weblog has a piece on starling syntax, which seemed relevant to my recent posting on simulations by birds, and recent reading on bots.

Hawks discusses recent work on bird song and experiments to see if birds can use sound to communicate in a linguistic sense. Not yet proven, seems to be his verdict.

On the other hand they can clearly use simulations to communicate – eg decoying a predator from the nest by simulating injury – so maybe whats at issue here is the nature of language. Reminds me of my earlier posting about the Turing Test: computers can simulate complex reality quite well, but meaningful language is another matter.

Also I found a fascinating paper yesterday about chatbots, a kind of AI system for providing FAQ answers on IRC chat channels – “they converse in natural language, they serve as a community memory, and they learn”. Here the problem is subtly different. No-one expects the bot to sound entirely like a human, so it does quite a good job of answering questions in a slightly robotic fashion. Instead, malicious users fill its poor head with junk information, which it duly repeats. (“Once people realize the bot is a bot, they often have a little fun with it, stuffing it with questionable information. The bots arent shy about expressing their newfound ideas in public, sometimes to the annoyance of channel curmudgeons.”)

It is very difficult to simulate human functioning completely. Ray Kurzweil would probably argue that we are getting nearer to notonly equalling but actually surpassing human responses; others like to believe that there is some sort of spark in the machine that will always resist mechanical copying. I simulate a human response by sitting on the fence here.

Leave a Reply

Your email address will not be published. Required fields are marked *