This isn't a black or white question either.

Most "ethical AI" companies are "fine-tuning" open source models with conventual data, but the training data of the base model itself is obfuscated.

You can't turn back time, but companies need to address this and be transparent about it.

Comments