My view is that LLM performance looks like bursts of improvement as breakthroughs happen, with lots of much smaller incremental improvements along the way. Wha

My view is that LLM performance looks like bursts of improvement as breakthroughs happen, with lots of much smaller incremental improvements along the way. What OpenAI were hoping for was AI stepping in to provide the breakthroughs. But for now it'll have to assist with incremental improvements only. There's probably 50% disappointment from the AI optimists and 50% relief from the pessimists.