18h agoDavid Pfau of Google DeepMind and creator Teortaxes argue LLMs merely repeat the status quo, making machine novelty an AGI-complete challengeBypassing dominant model outputs is likened to simple logit suppressionSentimentSentimentPos80%Neg20%Many users agree LLMs repeat the status quo and enthusiastically share prompting techniques for eliciting novel or post-human outputs, while one dismisses the approach as lying.10 comments with sentiment. View comments.