Project Updates: PartyMode, Simms AI
Published by KermMartian 16 years, 4 months ago (2007-11-21T01:30:35+00:00) | Discuss this article

I have been extremely busy with schoolwork and haven't posted any fresh news in quite a while, so I figured it might be time to talk about a few of the projects I've been working on lately. I'll talk first about PartyMode, a conglomeration of several projects including MSS 13.3, DiscoScreens, and VUfan. PartyMode is a network of systems designed to deploy an instant party in one of my school's computer/study labs at the press of a button. More specifically, it has a control panel with three switches and a large red modified Staples Easy button. The first red switch currently does nothing, but will eventually trigger the planned light effects (disco ball, perhaps?). The second red switch triggers an automotive actuator that pulls down all the light switches in the room. The third and final switch activates DiscoScreens, a C++ program that I wrote for Windows that displays a fullscreen color/brightness visualization of the currently-playing music based on network-streamed (UDP) data. Finally, the red Easy button queues a dance playlist in Winamp on the lab's media computer and begins playing it over MSS 13.3, the microLab 13.3 Surround Sound system. In addition, each of the red switches can trigger separate sound effect when turned either on or off (or both). The only current problems are some bouncing within the actuator protection circuitry that's making it switch off the lights at random times, and a small wiring accident that helped me set two of MSS 13.3's amplifiers on fire, leaving only 8.2 speakers operational. Hopefully, both of these issues will be repaired when I have time, most probably over my winter break when I'm finished with my first semester. I'll be documenting the various components of PartyMode as each is completed, and I will of course post about them here.

The second interesting project that I've picked up again is Simms AI (Artificial Intelligence), an experimental neural-networking learning system that I have built in bits and pieces. The current existing (and planned) components:

» Manager (Brainstem Manager): The manager is in charge of handling brain threads, modifying local output devices such as the 16x4 alphanumeric LCD on the side of Simms' physical shell, and killing off lagging brain thread (or restarting crashed threads). This section is almost completely functional at the moment, and generally succeeds in keeping Simms active for a current maximum of about two months without any administrative intervention.
» AIM (Instant Messaging Module): The AIM module connects Simms AI to the popular AIM instant messaging system under the screenname SimmsAI. This module is fully functional and currently online.
» Wiki (Wikipedia Correlation Module): The Wiki module constantly trolls through the popular online encyclopedia Wikipedia, loading a page at most once per thirty seconds based on popular keywords within its neural network, parsing the page, and adding the new data it has learned to its long term memory (LTM). This module is complete except for a possible residual bug.
» News (News Harvesting Module): Once every two minutes, the News module reads the top headline from the Google News RSS feed and files its content under the neural networking topic it decides is most pertinent from the body of the article. The intent of this module is to give Simms some basic idea of current events, perhaps even allowing it to correlate news over time. I have considered spawning an additional WorldTrends module to make Simms take on the task of calculating present and projected future data for my stagnant WorldTrends 3 project.
» Subconscious (Pattern-Matching/Consolidation Module): This module is the least complete within Simms AI's brain, meant to identify patterns and anomalies to add and remove connections, respectively. It runs in the background with no input or output interaction at all, and in function is quite similar to the human subconscious. I still need to code and debug most of this module.

From here on, I will be using whatever time I have to gradually improve the algorithms underlying Simms unilaterally, focusing particularly on the subconscious module. As usual, feel free to post comments or questions about either project.

Permalink

Advertisement