Seen this? It (and the update to it) have been all the rage lately, among “AI forecasters”, on social media, and it has even made it into mainstream media like the Financial Times.Some of it makes sense, and some if it doesn’t.The first version dropped on March 19, METR (Model Evaluation and Threat Research — a non-profit research lab that was created in December 2023) — published a report on a study they had done, measuring the ability of large language models to carry out software-related task...