We hope this will eventually enable us to diagnose failure modes, design fixes, and certify that models are safe for adoption by enterprises and society. It's much easier to tell if something is safe if you can understand how it works!
Understanding AI Models for Safety and Enterprise Adoption
By
–