Fascinating. A pressing question though: OpenAI and Google spend a lot of money (on low-wage workers) to train AI not to be so toxic and spew violent, sexist and racist remarks. Who does that work when AI is open source? Is it safe to assume this can be tackled via open source?

That's an important question. Projects like https://open-assistant.io have been attempting to use crowdsourced volunteer labor for that kind of tuning