← Jargon

Jargonmachine learning· 53 claims· 26 episodes

alignment problem

The unsolved challenge of shaping an AI's internal cognition so it reliably has the goals and virtues (such as honesty) that we intend, rather than merely appearing to.

As defined in gcs_chirp_transcripts_output_03ddb518_3b40_4b5a_aa80_27039507d7db_1781058711922_420_countdown_to_superintellig


Other definitions

Where it comes up · 26

and 18 more


Most important claims · 10 of 53Search all claims for “alignment problem” →