Outer Alignment
Outer alignment is the problem of specifying an objective, such as a reward function or training signal, that truly captures what humans want.
What is Outer Alignment?
If the objective itself is flawed, even a perfectly trained system will pursue the wrong thing.
How it works
Researchers improve objectives through human feedback, learned reward models and rules.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions