There is one big question in AI: how much better can LLMs get? This isn’t just important for worries about building a super-intelligent AI, it also matters for work, as AI can do many tasks at an average human level, or better, already. What happens if it reaches the 99% level of ability?

So far, LLM ability increases with the scale of the underlying model… and a founder of DeepMind says the scale of models is going up 1000x in the next 3 years!