DeepMind Is Now the Undisputed Leader in Language AI with Gopher (280B) | by Alberto Romero | Towards Data Science

Gopher’s bias and toxicity results prove once again how ubiquitous this issue is across LLMs and raise the same questions we’ve seen before: Is it worth building these models before we can remove their biased nature and toxic behaviors? Is it possible to do so without building the models in the first place? How can we de-bias AI correctly when doing so is a bias in itself? Do we want AI to reflect the world as it is or to help us reduce human-inherent discrimination?