near human-level speech recognition, open-sourced: t.co/eeuTfXJkwy (check out the examples, i find them difficult)

open-sourcing CLIP helped enable a lot of amazing work by the ML community, and i am hopeful the same thing will happen here!