I wrote an interface in ruby between the Twitter API and the chatterbox software MegaHAL. Why? I don’t know. Initially, I had it using the same brain developed over a few weeks of MegaHAL sitting in an IRC channel. I’ll write up the IRC wrapper later. This brain was a little too uncouth for my tastes, so I decided to start fresh for Twitter. The software gets 20 of the latest messages posted to Twitter every minute. It feeds them one-by-one into MegaHAL, and reads the reply. Due to the language it learned from IRC, the software isn’t answering @replies and isn’t posting updates anymore, it’s just learning. Once it begins to make good sense, I’ll let it post again. For now, you can view it’s progress live by clicking “more” below.
UPDATE: This project is no longer active. The MegaHal process that my wrapper code spawns began consuming a HUGE amount of CPU and memory. At this point the MegaHal process cannot load it’s dictionary within several hours. This was an interesting experiment and served it’s purpose, so I’ll let it rest as it is.
Nearly live data:
A “statement” is a tweet posted by a Twitter user. A “reply” is the response generated by MegaHAL.