That’s not necessarily wrong, but not the big explaining factor here I think. The technological challenges behind aligning ML models with factual reality aren’t solved, so it’s not an engineering decision. It’s more that AI is remarkably easy to market as being more capable than it is