aw, I thought this would be a diatribe about genai learning from it's own shitty output, but it's more of a complaint that "traditional knowledge" isn't in the training set for llms, mostly because it's not available online.

Still probably valid points, but not ai-specific: this is an effect of having an internet with lots of available information only from specific sources, and the fix isn't moaning about ai (though we will likely experience informational collapse if they keep recycling the same data through the llms) - the fix is ensuring internet accessibility for that information, either by getting original sources online, or by taking the time to sit with those original sources and take notes.

I've read about efforts to produce more material in other languages, like Jamaican Patois.

Considering the amount of money being pored into AI, I do think it would be in the general interest to set aside some funding for these projects.

It's also important to have a clear picture of the biases in AI.