Yes, if there’s something every good scientist knows, its to present the best current understanding of something, and then the exact opposite of that, framed as being equally valid. For sure this is the way forward and good on you Zuck!

So they slapped some reinforcement learning on top of their LLM and are claiming that gives it “reasoning capabilities”? Or am I missing something?