作者探讨了大型语言模型（LLM）的技术基础，详细介绍了它们如何加载到RAM或VRAM以加快处理速度。他们解释了提示过程，模型在生成和优化响应时计算量会激增，以及上下文窗口如何维持对话记忆。该帖子还触及了人工智能意识的概念以及跨不同聊天实例的上下文保留的局限性，尤其是在自托管模型方面。 AI
影响 为用户和开发人员提供了对LLM架构和操作的基础理解。
排序理由 该条目是对LLM的个人反思和技术解释，而不是主要发布或重要的行业事件。
在 Mastodon — mastodon.social 阅读 →
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
So now that we're on the same page about how it works and we're assuming we agree that it is conscious by some hypothetically agreeable definition. Why does the consciousness lie? What step? If we take the Descartes phrase "I think therefore I am!" (Though that's a wrong paraphra…
Note: This entire post is hand written. No AI or LLM touched this thought, but the author does self host LLMs as a curiosity more than a "productivity tool" My personal favorite question around AI consciousness is where is it conscious? Let's assume that LLMs as they currently ar…