Martin Dougiamas on Nostr: I had what feels like a good idea about #AI recently, which I'll blog about, but I ...

I had what feels like a good idea about #AI recently, which I'll blog about, but I wanted to share it here first.

What if we had a strong corpus of data about ethics, laws and so on, and the regulation/legislations that everyone keeps asking for made it illegal to release or use any AI models that did not include this corpus in its training data.

ie Core #curriculum for #AI.

I know it wouldn't solve everything, but surely it would at least give #LLMs a better chance of knowing right from wrong from the start.