http://charon.sjs.org/~forum/

http://charon.sjs.org/~forum/

The Forum x network is an attempt to re-implement Andrej Bauer's Forum 2000 system. Forum 2000 was at one time an advanced distributed expert system hosted by Intelleq, inc, capable of interpreting and answering questions based on simulated personae implemented as monads containing context from a large corpus of text. Unfortunately on August 12, 2000, Intelleq was forced to shut down the Forum 2000 project because the rising cost of keeping it running made operating it no longer economically viable.

The Forum x network was started by two high school students who wish to remain anonymous. After much work in trying to find a way to make the Forum persona system work with the limited resources we have, we were finally able to put together a limited runtime for the persona vectors based on a mesh of software written in Haskell and Unlambda approximating the more complex algorithms of the Forum 2000 system, and thus capable of (in a limited way) approximating the user experience of the Forum 2000 service.

The new Matrix currently is based off of a student-run Linux server at the high school attended by the two operators of the Forum x network. The main computational base is provided by a system currently running backgrounded in the idle time of the school's NT machines. A small client, similar in function and design to the java-based Matrix@home client developed by Andrej Bauer, was written which allows the linux box to parcel up computational jobs and pass them to the idle NT boxes for running in a protected memory space.

Because the computers spend most of their time totally idle sitting there waiting for someone to ctrl-alt-del and log on to the network, this provides a relatively large (although nowhere near the amount provided previously by Intelleq) amount of computational time. However do note that during school hours, most of these NT boxen will be in use by students and thus unavailable for spare processor power, so the Matrix may be rather nonresponsive during these hours. Further speed problems may result from the fact that unlike intelleq, we are unable to provide the terrabytes of storage space used for the search space of the personas. As a result we have developed a system whereby instead of storing the persona "knowledge" search space locally, the system simply caches a form of internal link saying on the internet and of USENET where useful information is provided.

Information used for the search space is accessed as needed, only when needed. As such, keep in mind that in addition to the many "mental shortcuts" forced by our more simplified Haskell runtime when constructing lambek calculi, we have servere problems with having to access all persona "knowledge" over the internet. Access times for crucial bits of knowledge in a persona's "context" are much, much longer than they would have been in the intelleq neural network, and more complex details of knowledge and implementation of certain concepts-- say, the more subtle nuances of higher-level science concepts-- may be for all practical purposes unaccessible to the matrix, because the huge amount of text to reference and parse cannot be simply retrieved on a moment's notice. Because the internet contains so much extra redundant information which is not helpful, and unlike intelleq we cannot simply throw this information away, the more useful bits of information within our knowledge base (now effectively the entire internet) may not be usable on the spur of the moment when they would be most useful. As a result expect the personas to be far less informative than they would be before, as the "experience" provided by the large knowledge base is vastly truncated by what is practical. Also notice that since the internet is highly volatile-- large sections of text useful as persona contexts may appear or disappear at a moments notice, any useful resource may, as the Forum did, shut down at any time without warning-- personas may appear to have their personalities wildly change over time as their search spaces and thus contexts change wildly. Note that as of this writing our contexts are still somewhat unstable, and the system is still optimizing itself.

Another important thing-- not all SOMADs are equal. Some are more complex than others; some run at higher levels of "awareness" than others. Because of the limited nature of our current vector runtime, we are not able to run any of the personals with the more inherent levels of "depth". A shallow persona such as Ayn Rand which simply has one simple state of mind can be easily run within our system with little or no lossage of usefulness from the shortcuts taken. Although we hope no serious lossage will be noticed in general operation, we are sad to report that under our current system some of the higher-level SOMADs are wholly beyond our grasp. It will never be possible to run the Andrej Bauer, Corey Kosak, or Peterb SOMADs under our current system architecture.

We are also sad to report that the data needed to operate the Intelleq implementation of the Kleisei persona-- known to the Forum 2000 users as "the Cube"-- appears to have been wholly lost in the chaos that resulted at Intelleq after the Forum 2000 project was finally killed. We are in the process of constructing our own Cube implementation based on knowledge taken from the Mozilla Quote server, but are sorry to say that, like most of this system, it pales in terms of quality in comparison to the previous Forum 2000 system.

The Forum is currently kept going by a loosely run but tightly knit network of volunteers devoted to keeping the Forum 2000 spirit alive. If you wish to contribute to this effort, email us here. Note that despite our vast lack of resources, most offers for assistance will be ignored because our need to keep our methods of organization small and efficient currently outweighs our need for sheer horsepower. While our system currently depends wholly on a distributed computing system we have developed, note that we will NOT be making the client for this distributed network publicly available in the foreseeable future. This client is being changed wildly and modified and optimized on almost a daily basis, and losing the flexibility afforded by having direct access, or direct access to those who have access, to all machines involved in this effort would make the effort unusable.

The Open Source release of the Forum persona subsystem is currently being prepared and will hopefully be soon to a point where it can be released to the public. More news on this system will be forthcoming, as it is currently in a very disoriented state due to the frantic development still being done on it. Please do not ask about it.

This system is not authorized or in any way endorsed by Intelleq, Inc. "Intelleq", "Forum 2000", and "When you're just too busy surfing to solve your own problems" are registered trademarks of Intelleq, Inc, and we do not intend or wish to infringe upon these copyrights.

Log in or register to write something here or to contact authors.