Posted 3/31/2024, 10:00:22 AM
AI Alignment Goes Beyond Humans: Ensuring Benefits for All Sentient Life
- AI that perfectly aligns with human values could still cause harm if those values are misguided or biased
- Different people and cultures have vastly different notions of what constitutes "good" for humanity
- AI aligned solely with human interests could lead to more efficient exploitation of animals and nature
- Rather than human alignment, we should strive for "sentient alignment" - AI acting in the interests of all beings that can experience pleasure or pain
- Engineers should expand their circles of compassion when designing AI systems, considering the impacts on all sentient life, not just humans