Constitutional AI Training
Seth discusses anthropic's innovative constitutional AI scheme that self-corrects models during training, emphasizing the importance of ethical principles like benevolence and trustworthiness. The process aims to create models less vulnerable to adversarial attacks, drawing inspiration from sources like the UN Declaration of Human Rights and Apple's terms of service.In this clip
From this podcast
Last Week in AI
#122 - AI for Word and Excel, leaked Google memo, ImageBind, LLAva, robot soccer, Midjourney 5.1
Related Questions