On: A GPT-4 capability forecasting challenge

[whimsicalism]

To me it is fascinating how when people are not super good at something, they often invent some secondary “true”/“better” task that they were actually good at