Trygve Kalland reposted
Glad more are realising this is a blunder! > By applying these tests outside their proper context [such as to LLM-based AI systems], the findings derived from such practice answer no real questions and fail to provide meaningful insights. doi.org/10.3384/nejl...