AI制品经理 | 入行AI的必须知识
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">随着大模型技术的快速发展，市面上涌现出了<span style="color: black;">海量</span>的大模型<span style="color: black;">制品</span>岗位，<span style="color: black;">那样</span>想要进入AI行业的<span style="color: black;">制品</span>经理<span style="color: black;">朋友</span>，需要提前做好<span style="color: black;">那些</span>准备工作呢？这篇<span style="color: black;">文案</span>里，作者总结了入行AI的<span style="color: black;">必须</span>知识，<span style="color: black;">包含</span>市场调研、<span style="color: black;">制品</span>底层<span style="color: black;">规律</span>等内容，<span style="color: black;">一块</span>来看。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/db5f2e4eb07f47d695b7b8a82d3511c0~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1727531291&x-signature=X8i93Htj%2FIHKGyANuj0Iz94zuVQ%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">AI大模型从去年11月<span style="color: black;">起始</span>到现如今，<span style="color: black;">已然</span>非常火热，无论大厂还是创业新星都在为AI的落地搭建<span style="color: black;">制品</span>架构，并<span style="color: black;">快速</span>跟进落地铺量，随着业务扩建，市面上<span style="color: black;">亦</span>涌现出<span style="color: black;">海量</span>的大模型<span style="color: black;">制品</span>岗位。<span style="color: black;">针对</span><span style="color: black;">日前</span>在看机会的<span style="color: black;">倘若</span>想要入AI坑的话，<span style="color: black;">制品</span>经理要<span style="color: black;">起始</span>做<span style="color: black;">那些</span>方面的准备工作呢？</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;"><span style="color: black;">1、</span>市场摸底调研：市面<span style="color: black;">重点</span><span style="color: black;">供给</span>AI服务都有<span style="color: black;">那些</span>大类？</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">选取</span>AI大模型的岗位本身<span style="color: black;">亦</span>是在做职业的规划过程，对线上已有的<span style="color: black;">机构</span><span style="color: black;">供给</span>的能力进行摸底调研有助于找准自己的兴趣点，并基于后续规划<span style="color: black;">选取</span>最为契合的岗位方向。<span style="color: black;">根据</span><span style="color: black;">制品</span>服务的类型，市面上AI<span style="color: black;">供给</span>的品类<span style="color: black;">重点</span>有</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">对话生成式</span></strong><span style="color: black;">：基于深度学习模型，如循环神经网络（RNN）或变种，如长短时记忆网络（LSTM）和<span style="color: black;">重视</span>力机制。这些模型<span style="color: black;">经过</span>学习<span style="color: black;">海量</span>的对话数据，<span style="color: black;">能够</span><span style="color: black;">捉捕</span>到语言的上下文和语义信息，并生成符合语法和语义规则的自然语言回复。这类<span style="color: black;">制品</span>在市面上比较多<span style="color: black;">亦</span>相对成熟，<span style="color: black;">例如</span><span style="color: black;">大众</span>熟知的openai、文心一言、glow等。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">AI绘图</span></strong><span style="color: black;">：利用人工智能技术进行绘图和创作的过程。<span style="color: black;">经过</span>训练深度学习模型，计算机<span style="color: black;">能够</span>学习并模仿艺术家的绘画风格、创作技巧和审美特点，从而生成<span style="color: black;">拥有</span>艺术性的图像和绘画作品。在应用这套技术的，<span style="color: black;">例如</span>抖音<span style="color: black;">制品</span>的特效、百度文心一言绘图功能。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">虚拟助手和智能客服</span></strong><span style="color: black;">：利用自然语言处理和对话生成技术，<span style="color: black;">能够</span>与用户进行语音或文本交互，<span style="color: black;">供给</span>信息<span style="color: black;">查找</span>、任务执行、问题解答等服务。这类<span style="color: black;">制品</span><span style="color: black;">一般</span>服务于2B比较多，<span style="color: black;">平常</span>在美团app里面的小美智能满足用户非结构化的找店<span style="color: black;">需要</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;"><span style="color: black;">陪同</span>类</span></strong><span style="color: black;">的<span style="color: black;">制品</span>市面上会有<span style="color: black;">有些</span><span style="color: black;">按照</span>人脸做面向分析，<span style="color: black;">亦</span>有<span style="color: black;">有些</span>心理咨询行业在用的咨询感情<span style="color: black;">陪同</span>类<span style="color: black;">制品</span>，底层都是基于用户在数据中的表现去匹配相应的情感支持。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;"><span style="color: black;">2、</span>这么多品类AI<span style="color: black;">制品</span>，它们底层<span style="color: black;">规律</span>是<span style="color: black;">怎样</span>实现的？</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">市面上已有的大模型<span style="color: black;">制品</span><span style="color: black;">重点</span>依赖底层的<span style="color: black;">基本</span>模型处理数据能力。从最底层到<span style="color: black;">实质</span>应用层，数据方向从<span style="color: black;">基本</span>数据的“非”智能转向有意图的“智能”。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/d937663231714b93a3f13a4837e4755a~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1727531291&x-signature=3z%2FLSxCUqezSC67nG3zO%2BWa8j4I%3D" style="width: 50%; margin-bottom: 20px;"></div>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">1. 基层模型能力</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">按<span style="color: black;">制品</span>所需能力建设所需的</span><strong style="color: blue;"><span style="color: black;">自研数据</span></strong><span style="color: black;">，<span style="color: black;">这儿</span>面可能<span style="color: black;">包含</span>特定的图像表达、语音表达、视频表达、<span style="color: black;">心情</span>感知等多种类型的数据服务能力。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">其次是基于</span><strong style="color: blue;"><span style="color: black;">战略合作的<span style="color: black;">特殊</span>数据源</span></strong><span style="color: black;">，本图中给出的数据合作<span style="color: black;">制品</span>ChatGLM-6B是开源的双语对话模型，含 62 亿参数，可处理对话聊天、智能问答等多种自然语言任务，支持在单张消费级显卡上推理<span style="color: black;">运用</span>，<span style="color: black;">供给</span>服务方是面向企业的2B类<span style="color: black;">机构</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">开放接口</span></strong><span style="color: black;">，众所周知，来自OpenAI跟微软Azure<span style="color: black;">供给</span>的智能化语义理解与智能办公的数据处理能力。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">开源模型</span></strong><span style="color: black;"><span style="color: black;">重点</span>分为NLP工具包、元学习开源库等资源信息，<span style="color: black;">根据</span><span style="color: black;">制品</span>本身的的<span style="color: black;">需要</span>能力<span style="color: black;">能够</span>自己接入。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">再往上走<span style="color: black;">便是</span></span><strong style="color: blue;"><span style="color: black;">基于<span style="color: black;">目的</span>服务的能力<span style="color: black;">怎样</span><span style="color: black;">保准</span>数据的<span style="color: black;">有效</span>调用</span></strong><span style="color: black;">，<span style="color: black;">通常</span>会从<span style="color: black;">安排</span>能力、推理优化、量化压缩几个象限去建构调用的资源，实现减少存储数据的压力，<span style="color: black;">提高</span><span style="color: black;">查找</span>速度，缩短问题被解答的思考时间，建构自适应学习能力<span style="color: black;">加强</span>问题解答满意度等<span style="color: black;">目的</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">最外层的输出，<span style="color: black;">便是</span>基于<span style="color: black;">以上</span>数据的</span><strong style="color: blue;"><span style="color: black;">安全合规</span></strong><span style="color: black;">，<span style="color: black;">通常</span>分为两个视角，</span><span style="color: black;">数据<span style="color: black;">运用</span>的安全监控能力、内容存储的合规合法能力</span><span style="color: black;">。关于数据隐私方面的内容<span style="color: black;">将来</span>会基于特定场景再做详述。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">2. 应用框架层</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">顾名思义，为了满足应用层的调取<span style="color: black;">有效</span>实现转换、理解、<span style="color: black;">查找</span>、调用、输出的架构层。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">经过</span><span style="color: black;">以上</span>架构图<span style="color: black;">亦</span><span style="color: black;">能够</span>看出，工具管理<span style="color: black;">重点</span>满足服务模型、Prompt、存储、知识图谱等模块的存放跟<span style="color: black;">运用</span>；</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">配置组合用于处理多个模型或组件组合在<span style="color: black;">一块</span>来实现更高级的功能。这种组合<span style="color: black;">能够</span>是串行的，即一个模型的输出<span style="color: black;">做为</span>下一个模型的输入；<span style="color: black;">亦</span><span style="color: black;">能够</span>是并行的，即多个模型<span style="color: black;">同期</span>处理<span style="color: black;">区别</span>的输入数据。<span style="color: black;">经过</span>组合<span style="color: black;">区别</span>的模型或组件，<span style="color: black;">能够</span>实现更<span style="color: black;">繁杂</span>的功能，例如语音识别和语义理解的组合、图像<span style="color: black;">归类</span>和<span style="color: black;">目的</span>检测的组合等。组合<span style="color: black;">能够</span><span style="color: black;">按照</span><span style="color: black;">详细</span>的<span style="color: black;">需要</span>和场景进行灵活的设计和<span style="color: black;">调节</span>，以实现更好的性能和效果。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">编排决策用来满足完成<span style="color: black;">繁杂</span>任务的执行。例如，一个语音识别应用可能需要<span style="color: black;">包含</span>语音前端处理、语音识别模型、语义理解模型等多个组件。在这种<span style="color: black;">状况</span>下，编排<span style="color: black;">便是</span>将这些组件<span style="color: black;">根据</span><span style="color: black;">必定</span>的<span style="color: black;">次序</span>和<span style="color: black;">规律</span>连接起来，以实现整体的功能。除了内容体裁的读取，编排还<span style="color: black;">触及</span>到模型的调用<span style="color: black;">次序</span>、输入输出的传递、数据的处理和转换等。它需要<span style="color: black;">思虑</span>到各个组件之间的依赖关系、数据流的管理、错误处理等方面的问题。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">编排技术在AI基层模型能力中起着重要的<span style="color: black;">功效</span>，它<span style="color: black;">能够</span><span style="color: black;">帮忙</span><span style="color: black;">研发</span>者更好地组织和管理<span style="color: black;">繁杂</span>的模型和组件，实现更<span style="color: black;">有效</span>、更灵活的AI应用。<span style="color: black;">同期</span>，编排<span style="color: black;">亦</span><span style="color: black;">能够</span><span style="color: black;">供给</span>更好的可扩展性和可<span style="color: black;">守护</span>性，使得AI模型能力在<span style="color: black;">区别</span>场景和<span style="color: black;">需要</span>下更加灵活和适应。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">以上的干货信息，<span style="color: black;">针对</span>想要入行的PM<span style="color: black;">来讲</span>面试基本上<span style="color: black;">已然</span>足够。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">3. <span style="color: black;">制品</span>应用层</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">这儿</span><span style="color: black;">重点</span><span style="color: black;">便是</span>市面上<span style="color: black;">大众</span>所关注到的各类<span style="color: black;">制品</span>，<span style="color: black;">详细</span><span style="color: black;">制品</span><span style="color: black;">这里</span>不<span style="color: black;">仔细</span>赘述。单独拎出来是<span style="color: black;">期盼</span><span style="color: black;">大众</span><span style="color: black;">能够</span>结合自己<span style="color: black;">平常</span><span style="color: black;">熟练</span>工作的视角来判断各类<span style="color: black;">制品</span>视角下，<span style="color: black;">咱们</span>的<span style="color: black;">优良</span>跟<span style="color: black;">能够</span>切入的<span style="color: black;">行业</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">例如</span>，<span style="color: black;">日前</span>在做客服平台，<span style="color: black;">怎样</span><span style="color: black;">才可</span>结合大模型<span style="color: black;">提高</span>整体的客服满意度？<span style="color: black;">倘若</span>在做<span style="color: black;">营销</span>管理，<span style="color: black;">怎样</span>利用好大模型更好的经营本地<span style="color: black;">营销</span>业务，这些问题想明白有助于投递大模型<span style="color: black;">关联</span>岗位中有的放矢，把完整的链路思考全面。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;"><span style="color: black;">3、</span><span style="color: black;">怎样</span>赢得市场增长跟变现思路</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">回复</span>这个问题本质是对变现模式的<span style="color: black;">回复</span>，在<span style="color: black;">这儿</span>抛砖引玉举两个例子。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">1. 自<span style="color: black;">媒介</span>从业人员</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">意见</span>从<span style="color: black;">运用</span>好AI<span style="color: black;">制品</span>，抓住AI生成<span style="color: black;">制品</span>能力的<span style="color: black;">优良</span>来落地，学习<span style="color: black;">运用</span>AI生成“显眼包”的<span style="color: black;">专题</span>，借用AI美化宣传文案，<span style="color: black;">乃至</span><span style="color: black;">能够</span><span style="color: black;">运用</span>AI润色图文，使用AI能力管理好社群<span style="color: black;">信息</span>分析运营思路。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">借用AI能力打磨账号<span style="color: black;">自己</span>的流量，<span style="color: black;">得到</span><span style="color: black;">必定</span>影响力，<span style="color: black;">连续</span>创造营收。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">2. 小企业<span style="color: black;">倘若</span><span style="color: black;">已然</span>有了相对成熟的体量且用户<span style="color: black;">类似</span>度较高，且不<span style="color: black;">期盼</span>只做<span style="color: black;">宣传</span></h1>付费订阅类<span style="color: black;">制品</span>：围绕<span style="color: black;">目的</span>群体，<span style="color: black;">供给</span>需要的<span style="color: black;">新闻</span>、动态、八卦等信息，整合输出高质量的内容社区，<span style="color: black;">供给</span>付费价值，收取订阅<span style="color: black;">花费</span>。对当前社会形势严峻的就业择业等问题<span style="color: black;">供给</span>信息<span style="color: black;">新闻</span>服务，向用户收取<span style="color: black;">花费</span>。数据<span style="color: black;">营销</span>和用户调研：借助大模型能力<span style="color: black;">经过</span>收集、分析和<span style="color: black;">营销</span>用户数据，为企业<span style="color: black;">供给</span>市场调研和用户洞察。借助AI能力，挖掘线上用户特征，输出用户本身所需的自我认知、潜能输出等的<span style="color: black;">制品</span>，实现平台变现通道。<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">以上，若对AI行业感兴趣的<span style="color: black;">朋友</span>欢迎来交流。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">专栏作家</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">蓝莲花zx，人人都是<span style="color: black;">制品</span>经理专栏作家。关注内容策略、内容后台、内容标签、账号策略等<span style="color: black;">行业</span>，<span style="color: black;">爱好</span>阅读，<span style="color: black;">期盼</span>做个有趣的人。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">本文原创发布于人人都是<span style="color: black;">制品</span>经理，未经许可，禁止转载</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">题图来自 Unsplash，基于 CC0 协议</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">该文观点仅<span style="color: black;">表率</span>作者<span style="color: black;">自己</span>，人人都是<span style="color: black;">制品</span>经理平台仅<span style="color: black;">供给</span>信息存储空间服务。</span></p>
你字句如珍珠，我珍藏这份情。 我完全赞同你的观点，思考很有深度。
页:
[1]