OpenAI更新模型准则，将“忠于人类意图”列为首项价值观Chinese only
OpenAI更新Model Spec，列出“忠于人类意图”、减少伤害、诚实透明、促进人的自主性和关怀五项核心价值观。新版准则强调，GPT应结合任务目的与对话上下文理解用户意图；涉及发邮件、删文件或花钱等现实行动时，须在明确授权范围内行事。
1 developments3 reports3 sources3 verified facts
Verified facts 3
OpenAI 更新 AI 模型规范，将忠于人类意图作为核心准则
OpenAI 表示，忠于人类意图是新版模型规范的基础，此次更新旨在让模型在陌生场景下践行人类价值观并接受人类监督；现有模型尚无法完全符合新版规范，团队正研究新的训练方式。
Event developments 1
OpenAI updates Model Spec with five core values
OpenAI’s updated Model Spec outlines five core values intended to guide more capable AI toward sound judgment, human agency, and faithfulness to human intent.
View source reports · 3 sources · 3 reports
All reports 3
OpenAI updates Model Spec, putting fidelity to human intent first
OpenAI has updated its Model Spec with five core values: fidelity to human intent, minimizing harm, honesty and transparency, supporting human agency, and care. The guidelines say GPT should consider a task’s purpose and conversational context when interpreting user intent, and stay within its explicit scope of authority when taking actions such as sending emails, deleting files, or spending money.
2 more sources reported this
- IT之家 科技新闻
- OpenAI 官方新闻
OpenAI updates Model Spec around five behavioral principles
OpenAI has reorganized its public Model Spec around five principles, including faithfulness to human intent, minimizing harm, and honesty and transparency. The company says models should understand the intent of users, developers, and OpenAI rather than follow instructions literally. It says current models do not yet fully meet the revised spec and that its team is researching new training methods.
2 more sources reported this
- 量子位 AI
- OpenAI 官方新闻
OpenAI updates Model Spec with five core values
OpenAI’s updated Model Spec outlines five core values intended to guide more capable AI toward sound judgment, human agency, and faithfulness to human intent.
2 more sources reported this
- 量子位 AI
- IT之家 科技新闻