4 ms·
Unintended behaviors in AI systems
- srikanthsrnvs 7y agoThe reinforcement learning ones are SO common. When I worked on autonomous tech a while ago, I remember constantly trying to tweak my reward function because of the agent learning that avoiding obstacles was a high enough reward in itself, so it just learned to go in circles constantly