@thatlev: The AI tests just got tested. Awkward. Epoch AI reviewed 15 benchmarks and labelled nine Flawed. Only four earned Verified, with two lacking enough information for a review. That doesn't mean every score is fake or every model is bad. It means the test behind that shiny leaderboard position deserves a closer look. Companies love announcing that their model passed the test. Slightly less convenient when the test is the thing getting called out #AI #AIBenchmarks #EpochAI #OpenAI #ArtificialIntelligence