Back to Lexicon
Wild WordAI Experience

alignment

Getting a model to pursue what humans actually want, not just what we said or rewarded.

First Seen

2010s

Context

The alignment problem asks: how do we ensure powerful AI systems act in accordance with human values and intentions? The challenge is that specifying what we truly want is harder than it sounds — systems optimizing a proxy reward may find unintended shortcuts. Became mainstream vocabulary during the GPT-4 era.

Citation

Alignment Forum / AI Safety

2022-24

https://en.wikipedia.org/wiki/AI_alignment

A wild word — documented from usage in the field, not coined by Xenolexica.

Ethics & Intention