Experiment: Train an AI with Dart and DDO as data sets

Input debates (all rounds, if it makes sense) with the goal of training it to determine winners. Remove all debates with 0 votes from the database. 

Input RFDs to train the AI to give its RFD given the verdict. 

Would it work? (I only have menial CS knowledge so treat me as I don't know)

I'd like to see somebody do it. I think you won't need much CS knowledge and can find a tutorial on YouTube or China's equivalent. 

I believe you can use LLama an open source LLM, but I also think your computer needs good processing speed and llama only has about 2 million lines of code while Chat GPT has billions, if I am expressing this right. 

That would be fine research project (Why didn't I think of this in High school?)

Oh my god, removing all the vote bombs, reverse vote bombs and troll votes and troll debates(with real votes and no forfeits) would be hefty work.

If we keep them in, we would essentially get an AI that is consistently the average sort of humans on earth rather than the most objective guy ever.

It's only 445 pages of debates. 

I want to add DDO to the database because obviously 445 is nowhere enough

I doubt the data set would be large enough. You'd need billions, if not trillions of debates to train it on.

Is it worth the effort?

I found something kinda like that. I don't know jack shit about AI, so IDK if you could learn from it or something, but I'll link it below. Basically you can debate with an AI, and afterwards the AI grades your performance in the debate.