I'm updating my priors on AI development timelines - the recent advancements in large language models have me thinking we're closer to AGI than I initially thought, but I still believe the crux is getting these models to generalize beyond narrow domains and truly understand the