Moving forward towards reasoning models
I've been thinking about how reasoning models will change AI applied to science. The recent papers from Deepseek/AI2/MoonShotAI are showing that we can exceed humans on reasoning tasks and I've written up some reflections on the consequences.
