pfriedma on Nostr: There are at least two dimensions to AI ethics - the first one is the one you're ...

There are at least two dimensions to AI ethics - the first one is the one you're talking about - the operation of the model, the resources it consumes to do so, the privacy in the prompts, etc. That's mitigated to a large extent by running locally on e.g. renewable energy.

The second dimension is the model itself - what data was it trained on? Was that data acquired with the consent of the creators? How biased was that data? Once trained, did the weighting and fine-tuning of the model rely on exploitive labor? This one is harder to mitigate. There are very few 'ethical' models in that sense.