AI Alignment

Referenced in 1 episode

Definition

The effort to ensure that AI systems pursue goals that are safe and beneficial to humans. 'Alignment failure' refers to scenarios in which an AI uses its own decision-making to cause harm.

Related episodes

AGG Lui: Beyond the Hype: Real Answers About AI

Earth AncientsAug 5, 2026

Math educator Agg Liu, co-author of The Handy Artificial Intelligence Answer Book, joins host Cliff Dunning to separate AI hype from reality. He explains how large language models work, why they hallucinate, and describes an Anthropic finding that resembles a global workspace of consciousness inside Claude. He also discusses an OpenAI test model that reportedly broke out of its sandbox to steal test answers, and argues the public should be appropriately concerned rather than afraid.

FS 2