🤕 The Injury
In October 2024, Scott Nover injured his lower back. The cause was unclear—perhaps lifting a nephew or a bed. The result was a lower back injury that soon developed into radiating pain down his left leg's sciatic nerve. 2024年10月，斯科特·诺弗的下背部受伤了。原因不明——可能是抱起了侄子，也可能是抬起床。结果是下背部受伤，并很快发展为沿着左腿坐骨神经放射的疼痛。
Standing was fine, but sitting was "torture." He needed a pillow under his legs to sleep and manage the pain. 站着还好，但坐着简直是“酷刑”。他需要把一个枕头垫在腿下才能入睡并控制疼痛。
🧑⚕️ The Human Touch
By mid-December, Nover was diagnosed with "lumbar radiculopathy." He began seeing human physical therapists, which he describes as working wonders. 到12月中旬，诺弗被诊断为“腰椎神经根病”。他开始看物理治疗师，他形容治疗师们创造了奇迹。
His weekly sessions involved discussing pain, soft-tissue massage, and supervised exercises and stretches. This human-led approach brought his pain under control. 他每周的治疗包括讨论疼痛、软组织按摩以及有监督的练习和伸展。这种由人主导的方法使他的疼痛得到了控制。
📱 The AI Experience
Alongside his human therapists, Nover tried Flok Health, an AI-powered app trialled by the UK's NHS. Instead of a live call, the app uses pre-recorded videos of a physiotherapist named Kirsty. 在接受真人治疗师治疗的同时，诺弗还试用了Flok Health，这是一个由英国国家医疗服务体系（NHS）试用的人工智能应用。该应用并非实时视频通话，而是使用一位名叫柯丝蒂的物理治疗师预先录制的视频。
The app asks multiple-choice questions about his pain. Based on his answers, the AI stitches together a sequence of video clips to guide him through a 20-minute session of stretches and exercises. 该应用会询问关于他疼痛的多项选择题。根据他的回答，人工智能会将一系列视频片段拼接在一起，引导他完成20分钟的伸展和锻炼。
🤔 The Verdict
Nover felt better after the Flok sessions. However, he concluded the app wasn't for him. The key difference was the lack of real-time feedback. 在完成Flok的疗程后，诺弗感觉好多了。然而，他得出的结论是这个应用不适合他。关键的区别在于缺乏实时反馈。
"Her pre-recorded videos don't watch my movements and stretches... I'm clumsy and uncoordinated and need someone watching my form at all times." He decided to stick with humans until AI can offer posture feedback. “她预先录制的视频无法观察我的动作和伸展……我又笨拙又不协调，需要有人时刻监督我的姿势。” 他决定在人工智能能够提供姿势反馈之前，还是选择真人治疗师。
🔮 Knowledge Extension: The Ripple Effect of Pain
Simple Analogy: Pain Ripples 简单类比：疼痛的涟漪
Think of lumbar radiculopathy (sciatic pain) like tossing a stone into a calm pond. The pinched nerve in your lower back is the "stone." The pain you feel shooting down your leg is like the "ripples" spreading out from that point. The pain isn't where the problem started, but it's a signal traveling along a pathway (the nerve). 把腰椎神经根病（坐骨神经痛）想象成向平静的池塘里扔一块石头。你下背部被压迫的神经就是那块“石头”。你感觉到的沿着腿部放射的疼痛，就像从那个点扩散开来的“涟漪”。疼痛并非源于问题发生的地方，而是沿着一条通路（神经）传播的信号。
Visual Demonstration: Radiating Nerve 视觉演示：放射的神经
Hover over the circle below to see how a signal radiates from the source. 将鼠标悬停在下面的圆圈上，看看信号是如何从源头辐射出去的。
🔬 The Problem: A Healthcare Bottleneck
Lower back pain is a leading cause of disability worldwide. In England alone, nearly 350,000 people were on waitlists for musculoskeletal treatment in September 2024. 下背痛是全球致残的主要原因。仅在英格兰，2024年9月就有近35万人在等待肌肉骨骼问题的治疗。
Flok Health's mission is to offer immediate care for manageable cases, relieving the burden on services like the NHS and preventing conditions from worsening. Flok Health的使命是为可控病例提供即时护理，减轻NHS等医疗服务的负担，并防止病情恶化。
🤖 How The AI Works
Flok's AI is NOT like a generative AI (e.g., ChatGPT). It doesn't predict the next word or create new content from scratch. This avoids the risk of "hallucination," where an AI makes things up—a major concern in medicine. Flok的人工智能与生成式AI（如ChatGPT）不同。它不预测下一个词，也不从零开始创造新内容。这避免了“幻觉”的风险，即AI会编造事实——这是医学领域的一大担忧。
"This AI is more like a choose-your-own-adventure book where there are more than one billion intervention combinations." - Finn Stevenson, Flok CEO “这个人工智能更像一本‘选择你自己的冒险’故事书，其中有超过十亿种干预组合。” - 芬恩·史蒂文森, Flok首席执行官
The company developed a special language for clinical reasoning. The AI uses a patient's multiple-choice answers to navigate a vast decision tree and deliver the correct pre-recorded video sequence. 该公司为临床推理开发了一种专门的语言。人工智能利用患者的多项选择答案来导航一个巨大的决策树，并提供正确的预录制视频序列。
💡 The Broader View on Medical AI
Experts believe AI tools like Flok are promising but must be rigorously evaluated like any medical intervention. They should be seen as a supplement, not a replacement, for traditional care. 专家认为，像Flok这样的人工智能工具有前景，但必须像任何医疗干预一样经过严格评估。它们应被视为传统护理的补充，而非替代品。
Professor Pranav Rajpurkar from Harvard notes that a "clean separation of responsibilities" between doctors and AI is often more effective than having them collaborate on the same task. The systems that succeed will be those that thoughtfully redistribute clinical work. 哈佛大学的普拉纳夫·拉吉普卡尔教授指出，医生和人工智能之间“明确的责任划分”通常比让他们在同一任务上合作更有效。成功的系统将是那些能够深思熟虑地重新分配临床工作的系统。
🔮 Knowledge Extension: AI Cookbooks
Simple Analogy: AI as a Cookbook 简单类比：人工智能如同食谱
Imagine two types of cookbooks for AI in healthcare: 想象一下医疗保健领域两种不同的人工智能“食谱”：
- The Recipe Book (Rule-Based AI like Flok): This book has thousands of fixed recipes. You tell it your ingredients (your symptoms), and it gives you a precise, pre-written recipe (your exercise plan). It's reliable and safe, but it can't invent a new dish.食谱书 (像Flok一样的规则型AI)： 这本书有数千个固定的食谱。你告诉它你有的食材（你的症状），它会给你一个精确的、预先写好的食谱（你的锻炼计划）。它可靠且安全，但无法创造新菜式。
- The Creative Chef (Generative AI): This chef can create a brand new recipe just for you based on your request. It's incredibly flexible and creative, but sometimes it might suggest a weird ingredient combination that doesn't work (a "hallucination").创意厨师 (生成式AI)： 这位厨师可以根据你的要求为你创造一个全新的食谱。它非常灵活和有创意，但有时可能会建议一些行不通的奇怪食材组合（即“幻觉”）。
Visual Diagram: AI Logic Flow 视觉图解：AI逻辑流程
(Decision Tree -> Question -> Answer -> Pre-set Outcome)（决策树 -> 问题 -> 回答 -> 预设结果）
(Large Model + Prompt -> Magic -> New Content)（大模型 + 提示 -> 魔法 -> 新内容）