"Given this is based on a non-peer-reviewed paper by researchers at Anthropic, with their track record on research we can probably assume a) that the methodology is really bad, and b) if the results are accurate (if they even test the thing purported to be studied), they think this is good PR for AI"
A little bonus pastoral from my drive through the outskirts of Klamath County yesterday. We had heavy rain, then a flash of rainy sun, and even a bit of snow up in the mountains.