Skip to content

Model Training & Adaptation

Proximal policy optimization

A way of teaching AI through rewards while discouraging changes that are too large during learning.

Example

AI learns rewarded behavior while limits discourage very large learning changes.

Why people use it

It helps an agent improve without changing its behavior too sharply at once.

What you'll hear

“Improve the choices without making the learning unstable.”

What this means for you

Check behavior beyond the activities that earned rewards during practice.

Can you control it?

Developer-only

The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.

Common questions

Does PPO guarantee safe or desirable behavior?
No. Results depend on the reward, data and training process.
Does it require human scores for every action?
No. Feedback can come from a task or another scoring process.
Can a higher reward hide a worse real-world result?
Yes. A poorly chosen reward may favor behavior that people do not actually want.

Related terms

Still have questions?

Up to 500 characters.

Ask LATHIC about AI. Relevant glossary entries may be included.

Your question, the glossary entries it matches, and a rotating pseudonymous identifier go to Microsoft Azure’s OpenAI service through Vercel AI Gateway to generate an answer. Zero retention and no training are required of the provider, and LATHIC does not save your question or answer. Privacy Notice