Administered by:
People impressed that some new LLM outperforms humans at IQ tests
JFC there's so much to unpack here but let's just yeet the whole suitcase into the canal
@pikesley a definite yikes there this morning. Let’s automate the taking of exams rather than reconsidering whether regurgitating raw information from memory is the way to assess ability in a world with Google.
@jon Bruce nails it https://functional.cafe/@otfrom/112048185988327223
Functional CaféChip Butty (@otfrom@functional.cafe)@pikesley@mastodon.me.uk I do love how LLMs mathematically prove how much is wrong with how we do things atm (they show bias b/c we are biased, they do well on things like IQ b/c IQ is a terrible measure)
@jon moderately concerned that we have an Actual True Believer in our midst