Reinforcement Learning
A training method where a model is rewarded for good outputs and penalized for bad ones, so it learns preferred behavior.
It is a big reason modern assistants feel helpful and polite rather than raw and random.
Frequently asked questions
Why do modern AI assistants feel so polite and helpful?
A big reason is reinforcement learning, where the model is rewarded for good outputs and penalized for bad ones until it learns the behavior people prefer. Without it, a raw model tends to feel erratic and unfiltered.
How is reinforcement learning different from regular training?
Regular training teaches a model to predict language from huge amounts of text, while reinforcement learning shapes its behavior afterward through rewards and penalties. It is less about learning facts and more about learning which responses are wanted.
Does reinforcement learning make an AI trustworthy?
It helps steer behavior toward being helpful, but it is one part of the larger work of alignment, not a full guarantee. A model can still make mistakes or hallucinate, so human review of important outputs still matters.
New to all this? Start with what an AI agent really is, browse the full glossary, or explore the learning hub.
← Back to the glossary