语言学家 Emily Bender 认为，尽管大型语言模型 (LLM) 在转录和翻译等任务中有用，但目前的聊天机器人界面并非有益。她批评 LLM 使用人称代词的设计选择，这会产生自我和意识的错觉。一项关于 AI 在放射学领域应用的研究表明，LLM 缺乏真正的推理能力，只是生成听起来貌似有理的文本，而不理解真假等概念，这对于依赖严格证据的医学等领域来说令人担忧。 AI
影响 引发了对 LLM 推理能力固有局限性以及欺骗性界面的潜在担忧，影响了对 AI 应用的信任，尤其是在医学等关键领域。
排序理由 该集群包含来自知名专家 (Emily Bender) 对 LLM 性质和应用的批判性意见和分析。
在 Mastodon — mastodon.social 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
Another good quote: In Bender’s view, there is simply no good way to interact with an LLM. Language models can be “extremely useful for things like automatic transcription and machine translation. There's a role for that kind of technology,” she says. “But turning it around into …
https://www. the-independent.com/tech/ai-ne ws-humanize-chatbot-conscious-b2963788.html Worth reading just for quotes from Emily Bender. Here's a quote about one of my pet peeves: “The fact that these systems output ‘I/me’ pronouns is totally a design choice,” [Emily Bender] says…
@ thomasfuchs @ emilymbender This article is about a study of the use of # LLM # AI in radiology and other medical areas. The study’s conclusions confirm what we already know: LLMs do not (cannot) reason. They sound plausible, but they will say anything so long as it seems plausi…