In a way, this seems like sour grapes to me. I bet if it was their own company/institution that released gpt-4, they wouldn't be calling for this pause.
Actually one of the founders of the "Future of life" institute that organized this is a DeepMind "AI safety" employee Victoria Krakovna (who also signed it), and I'm guessing she'd not be doing this without DeepMind's approval. DeepMind's whole raison d'etre was of course building AGI, which they were initially pursuing via RL before belatedly joining the LLM bandwagon. I'm sure they'd welcome a 6-month catch-up period!
