>>714644957
>The goal is to scale them up until they're useful enough to help develop new AI systems
Again, a fool's errand: LLMs can only ever understand its own spoonfed data. It will never, ever be fucking creative. This isn't a debate, retard. It's a fact.