Скачать или смотреть Tim Parker — Formalizing Ethics In Logic [TAIS 2024]

In this presentation I will present a framework for ensuring safe and ethical behaviour in AI by using values written in formal logic, while considering various practical and philosophical issues regarding the formalising of ethical values. A prioritised value base can be used to verifiably guarantee that certain behaviours will always be avoided. Additionally, I will introduce various notions of responsibility anticipation, meaning future responsibility for value violation can be used to compare plans. This ensures that safety is maximised even in scenarios where a safe outcome cannot be guaranteed, when dealing with unpredictable agents, or when multiple concerns must be addressed. I will also show how the logical notion of Seeing To It That (STIT) can be used to balance safety concerns with respect for human autonomy.

This is a recording from TAIS 2024, a technical AI safety conference hosted at the Plaza Heisei in Tokyo April 5th–6th. TAIS 2024 was organised by AI Safety Tokyo, sponsored by Noeon Research, in collaboration with AI Alignment Network, AI Industry Foundation and Reaktor Japan.

0:00 Talk
17:54 Q&A