Model Training & Adaptation
Value function
An estimate of how much future reward an AI can expect from a situation or action.
Example
An agent estimates how promising a game position is.
Why people use it
It helps an agent compare how promising different situations or choices seem.
What you'll hear
“How much future reward might follow from here?”
What this means for you
Identify the reward definition before interpreting a high estimated value.
Can you control it?
No
No direct control. This describes a wider issue, concept or result rather than something you can simply switch on or off in a tool.
Common questions
- Does value here mean moral worth?
- No. It refers to expected reward under the task's scoring rules, not a judgment about ethics.
- Can the highest estimated value be wrong?
- Yes. It is a learned estimate and can be inaccurate in unfamiliar situations.
- Does it only count the next reward?
- Not necessarily. It can consider rewards expected further into the future.