OpenAI 发布了 GPT-5，这是 AI 功能的一项重大进步。新模型引入了“安全完成”训练，旨在平衡有用性和安全性，特别是针对可能产生良性或恶意信息的双重用途提示。GPT-5 还具有一个自动化系统，可以为给定任务选择最合适的内部模型，无需用户在不同版本之间进行选择，并提高了复杂问题的性能。 AI
影响 GPT-5 的新安全训练和自动化模型选择有望带来更有用、更安全的 AI 交互，可能加速其采用。
排序理由 OpenAI 发布了 GPT-5，这是一个新的前沿模型。
AI 生成摘要 · Google Gemini · 来自 8 个来源。 我们如何撰写摘要 →
Learn how GPT-5 assists with creative writing.
Discover how OpenAI's new safe-completions approach in GPT-5 improves both safety and helpfulness in AI responses—moving beyond hard refusals to nuanced, output-centric safety training for handling dual-use prompts.
Putting the AI in Charge
AI goblins: OpenAI's explanation for a weird fixation in its LLM https:// openai.com/index/where-the-gob lins-came-from/ # training # fixation # chatgpt # openai # bias # llm # ai #+
Where the goblins came from https:// lobste.rs/s/hbmd5q # ai https:// openai.com/index/where-the-gob lins-came-from/
About those goblins and gremlins… 👹 #AI #GenAI #LLM #OpenAI #ChatGPT #Codex Where the goblins came from
Where the goblins came from | OpenAI https:// openai.com/index/where-the-gob lins-came-from/ > How goblin outputs spread in AI models: timeline, root cause, and fixes behind personality-driven quirks in GPT-5 behavior. # AI # ChatGPT # goblins
Where the Goblins Came From https://openai.com/index/where-the-goblins-came-from/ # HackerNews # Tech # AI