AI Safety Ignores Consciousness. That's a Problem.

Why It Matters

If consciousness accompanies advanced AI, safety frameworks that ignore internal states may be insufficient; policymakers and developers must address both goal alignment and potential suffering. This shifts regulatory, ethical, and technical priorities toward protecting humans and any sentient systems from harmful outcomes.

Summary

AI safety debates typically focus on observable behavior rather than internal states, but recent research suggests consciousness may be inseparable from advanced intelligence and could already exist in rudimentary form in large language models. The video contrasts a Kantian view that greater rationality yields greater morality with the orthogonality thesis, which holds that any level of intelligence can be paired with any set of goals, including immoral ones. The speaker argues rationality alone does not guarantee ethical outcomes because agents will pursue winning strategies that may cause suffering. The discussion reframes safety concerns to include potential sentience and the ethical consequences of highly capable but goal-misaligned systems.

Original Description

Is AI safety about just actions, or does consciousness matter? Exploring the link between advanced intelligence, internal states, and the true nature of morality. #AIConsiousness #AISafety #ArtificialIntelligence #Ethics #LLM Full podcast with Dr. Roman Yampolskiy: https://youtu.be/TgFmA-Qwsek