> Makes me wonder about using the bee-l archive (and perhaps other forums) to train the LLMs (large language models).

I've worked in this area of inquiry since the days when we thought Prolog was a good way to make "AI", and endless debates between "PL/1" and "PL/M", two now-defunct languages, occupied most of Comp.Sci.Languages on UseNet, and spilled over into a completely separate BitNet-only group that Yale started.

Neural nets and other AI tools work well in some areas, as "training" the AI can lead humans, who perceive only some patterns, to see others.  An example of how effective this is would be the natural growth a slime mold duplicating the design work for the Tokyo Subway.

But Bee-L would not a good source of data, as contradictions are many, most arguments offered are specious, and rigorous arguments are rarely consider "polite" but those who refuse to accept that, as I and others have pointed out many times in multiple ways with real-world examples, "science is a full-body contact sport".

So, what actual knowledge would a LLM gain by reading Bee-L?  Only that "some people say X" while "other people say NOT X".  

The only uncontradicted statement made in the entire history of Bee-L, including the posts not backed up and lost from the early pre Aaron Morris (Ed Southwick/BitNet) days would be Pooh's, in the first A.A. Milne Pooh book:

"They might or they might not,” said Winnie-the-Pooh. “You never can tell with bees.” He thought for a moment and said: “I shall try to look like a small black cloud. That will deceive them.”

If we would rather not have our thoughts harvested and mangled by an AI, it would cost under $50 to copyright the entire archive in the USA, and then announce that the entire archive is only available under a "Creative Commons Attribution-NonCommercial-NoDerivs (CC-BY-NC-ND)" license, which prohibits commercial exploitation, but allows non-profit use.  The big task would be asking L-Soft to dump us out a complete copy of the archive in flat ASCII, as one must submit "a copy of the work" with the application to the LOC.  Such applications are invariably approved, unless there is an active lawsuit involving the works to be copyrighted.

             ***********************************************
The BEE-L mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software.  For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html