王峰 AI × 经营 约一次诊断

总目 / 文章 / WF-0435

文章 · 存档

一天生成6000个广告,难道AI就只是用来高效忽悠人类?

看到“一天生成6000个广告”这种消息,心里跟吃了个苍蝇一样难受。人类折腾了这么多年,烧了这么多钱,堆了这么多算力,最后最先跑通的赚钱路径,居然是把广告做得更快、更密、更准,然后更高效地去围堵人的眼睛和耳朵。你刷手机,它追着你;你打开视频,它贴着你;你刚起一点欲望,它立刻替你把欲望包装成“立即下单”

一天6000个广告真正值钱的不是创意,是筛选;真正被研究的不是产品,是人性漏洞;真正被优化的不是内容质量,是刺激反应。谁还在意一句话有没有灵魂?只要能让你手抖一下点进去,它就算完成任务。

这就把AI当前最擅长的那一面暴露得很彻底。它厉害,确实厉害,但它的厉害高度依赖一个前提:工作结果可以接受概率,过程可以接受试错,内容可以接受大规模复制,错误成本又不会立刻炸穿天花板。广告刚好全中。标题可以试,图可以试,文案可以试,视频口播可以试,今天不行明天再换一批,反正系统会替你筛。它不需要真正理解人,只需要不断逼近人的反应;它也不需要真正有思想,只需要在海量已有内容里拼出更容易让人点开的那个版本。这种活儿,AI做起来当然凶猛,因为这类活本身就不是在追求“唯一正确”,而是在追求“高概率有效”。

所以我对AI有个很不浪漫的判断。它先吃掉的,从来都是那些本来就不太需要动脑子的工作。有人害怕AI取代作家、音乐家、画家、科学家,听上去很唬人,仿佛人类文明马上要被机器按在地上摩擦。可真往深处看,这些职业里最硬的部分,AI根本碰不到。它能模仿文风,却碰不到真正的洞察;它能拼装旋律,却碰不到真正的情感经验;它能画出漂亮图像,却碰不到一个创作者为何非要这样构图、为何在这里停顿、为何在那个细节上狠下心;它能整理论文资料,却碰不到那个把旧世界捅出一个新洞的念头。创造这件事,本来就不是把旧东西排列得更像样一点。创造是人类从没有过的那一下,是世界原本没有,后来被某个人硬生生弄出来的那一下。

AI更像什么?更像一个规模化的熟练工,一个极其勤奋、极其会抄、极其会归纳、极其会试错的熟练工。只要人类做过,只要样本足够多,只要目标大致清楚,它就能卷得你怀疑人生。于是写手危险了,因为流水线文字本来就没多少思想;抄谱的人危险了,因为照着规范转写原本就是机械劳动;画工危险了,因为只负责执行、不负责判断的那部分最容易被吞掉;实验员里那些重复操作、记录、整理的人也危险了,因为流程一旦标准化,机器天然占优。这个逻辑放到财务会计上更残酷。财务会计里90%的工作本来就是在重复、校验、填制、搬运、对账、归档、套格式、跑流程。嘴上说专业,手上干的却常常是体力活。这样的工作,不被AI替代才奇怪。

财务人听到这里会不舒服,觉得这话刺耳。可刺耳不代表不真。月结流程里有多少动作只是机械重复?报表分析里有多少内容只是把数字换个版式再发一次?预算编制里有多少表是填了没人看、看了没人用?税务、成本、核算、往来、报销、合同归档,里面又有多少环节只是人坐在那里充当一个会点击鼠标的接口?如果这些年一直把“会干活”等同于“会复制过去的动作”,那AI一来,当然先冲着这一层下手。机器不累,不烦,不要加班费,也不需要情绪价值,它最适合接这种活。

真正有分量的问题,从来不是“AI会不会替代人”,而是“你做的工作里,到底有多少部分值得保留给人”。这个问题一问,很多岗位就开始发虚了。因为太多人平时并没有真正用心工作。所谓用心,不是加班更晚,不是表格更整齐,也不是汇报更流畅。用心是你有没有判断,有没有穿透表象看到问题的能力,有没有把不同信息串成逻辑链条的能力,有没有在混乱里抓住关键矛盾的能力,有没有在责任要落下来时真的敢拍板。AI替不了这种东西。它可以帮你算,帮你找,帮你比,帮你生成,帮你预警,甚至帮你写出一版像模像样的初稿,但最后那个“为什么是这个答案”“代价会落在哪儿”“这条路值不值得走”,还是得人来扛。

所以今天最讽刺的画面,一边是有人用AI一天制造6000个广告,拼命把技术变成收割注意力的喇叭;另一边是大量真正需要AI的岗位,还在拿它写点空洞文章、做点花哨PPT、画点看起来很聪明的流程图。一个在过度使用,一个在低水平使用,最后大家都在喊AI改变世界,真正被改变的却只是点击率和焦虑感。技术没有错,问题出在人类总喜欢把最快见钱的地方当成“最重要的方向”。广告行业先赚到这笔钱,不代表这就是AI最体面的用途,只能说明商业世界对“收割”永远比“建设”更敏感。

我反而越来越相信,越是认真工作的人,越该主动掌控AI。因为AI真正值钱的地方,从来不是替你表演,而是替你减掉那些不值得你消耗生命的烂活。它该去做初筛、初算、初稿、初步归类、流程推动、重复核对、结构整理,让人把时间拿回来,去做判断、做设计、做创造、做沟通、做博弈、做决策。它该成为得力助手,不能被供成电子神明。谁把它当工具,谁会越来越强;谁把它当脑子,谁迟早把自己的脑子养废。

一天6000个广告当然能赚钱,也当然足够吓人。可这种“成功”越热闹越让人警惕。它提醒的不是AI有多伟大,而是人类多容易把伟大的东西先拿去做最短视的事。真想让AI有点体面,先别忙着让它替你忽悠别人,先让它替你把工作做得更扎实一点,把脑子解放出来,把真正该由人完成的那部分,做得更像人。

Is AI only for fooling people better?

If AI can make 6,000 ads in one day, is AI only for fooling people better?

When I see news like “6,000 ads in one day,” I feel sick. People have worked on this for so many years, spent so much money, and used so much computing power. But in the end, the first clear way to make money is to make ads faster, more, and more exact, and then use them to catch people's eyes and ears better. You scroll on your phone, and it follows you. You open a video, and it sticks to you. The moment you feel a little want, it quickly wraps that want into “buy now.” The more I look at it, the more strange it feels. AI is so strong, but it is first used to push people to spend money, pull people into quick choices, and make human weak points bigger. Tech is flying forward, but human life may not move forward with it. That gap really feels like a high-level world ending up selling loudspeakers on the street.

The truly valuable part of 6,000 ads in one day is not ideas. It is picking. What people really study is not the product. It is the weak holes in human nature. What is really being improved is not content quality. It is how to cause a fast reaction. Who still cares if one line has a soul? If it can make your hand shake and click, then the job is done.

This shows very clearly the side of AI that is strongest right now. It is strong, truly strong. But its strength depends a lot on one thing: the result can be based on chance, the process can allow trial and error, the content can be copied on a huge scale, and the cost of mistakes will not blow up the roof right away. Ads fit this very well. Titles can be tested. Pictures can be tested. Copy can be tested. Video words can be tested. If today does not work, tomorrow you change a new group. Anyway, the system will do the picking for you. It does not need to truly understand people. It only needs to get closer and closer to people's reactions. It also does not need to truly have ideas. It only needs to use a huge amount of old content to build the version that gets more clicks. Of course AI is fierce at this kind of work, because this kind of work is not trying to find “the one right answer.” It is trying to find “what will most likely work.”

So I have a very plain view of AI. The first jobs it will eat are always the jobs that did not need much deep thinking in the first place. Some people fear AI will replace writers, musicians, painters, and scientists. It sounds scary, like human civilization will soon be pushed to the ground by machines. But if you look deeper, AI cannot touch the hardest part of these jobs at all. It can copy a writing style, but it cannot touch real insight. It can put together a tune, but it cannot touch real feeling from life. It can draw pretty pictures, but it cannot touch why a maker had to use this shape, why the maker stopped at this point, and why the maker was so hard on that one detail. It can sort papers and study notes, but it cannot touch that one thought that cuts a new hole in the old world. Creation was never just making old things look a little better in a new order. Creation is that one move that never existed before, and then one person brought it into the world by force.

So what is AI more like? It is more like a skilled worker at huge scale, a worker who is very hard-working, very good at copying, very good at sorting, and very good at trying again and again. As long as humans have done it before, as long as there are enough examples, and as long as the goal is mostly clear, it can work so hard that it makes you doubt your life. Then writers are in danger, because factory-line words never had much thought in them. People who copy music notes are in danger, because changing old notes into the right form was already mechanical work. Art workers are in danger, because the part that only carries out orders and does not make judgment is the easiest to swallow. In labs, the people who do repeated work, write things down, and sort things are also in danger, because once the process is standard, the machine is naturally better. This logic is even more hard in finance and accounting. In finance and accounting, 90% of the work is repeated work, checking, filling forms, moving data, checking accounts, saving files, using set formats, and running steps. People say it is professional work, but often the hands are doing body work. If work like this is not replaced by AI, that would be strange.

Finance people may feel bad when they hear this. They may think these words hurt. But words that hurt are not always false. In month-end work, how many steps are only repeated work? In report analysis, how much is only taking the same numbers and sending them again in a new look? In budgeting, how many sheets are filled out, but no one reads them, and even if someone reads them, no one uses them? In tax work, cost work, accounting work, back-and-forth account work, payment back work, and contract filing, how many steps are just a person sitting there and acting like a clicking interface? If for many years people kept saying “can do work” when they really meant “can repeat old moves,” then when AI comes, of course it will hit this layer first. Machines do not get tired. They do not get annoyed. They do not ask for overtime pay. They do not need emotional comfort. This kind of work is exactly what they fit best.

The truly heavy question was never “Will AI replace people?” The real question is “How much of your work is really worth keeping for people?” Once you ask that, many jobs start to feel weak. Why? Because too many people were not truly working with their full mind before. “Full mind” does not mean staying later at work. It does not mean cleaner tables. It does not mean smoother reports. It means this: do you have judgment? Can you see through the surface and find the real problem? Can you join different pieces of information into one chain of logic? Can you catch the key problem inside a mess? When duty falls down on someone, do you really dare to make the call? AI cannot replace that. It can help you count, help you find, help you compare, help you make text, help you warn, and even help you write a first draft that looks pretty good. But in the end, the question “Why is this the answer?”, “Where will the cost fall?”, and “Is this road worth taking?” still has to be carried by a person.

So the most sharp picture today is this. On one side, people use AI to make 6,000 ads in one day and turn tech into a loudspeaker that cuts and grabs attention. On the other side, many jobs that truly need AI are still using it only to write empty articles, make flashy PPTs, and draw flow charts that look smart. One side is using it too much. The other side is using it at a low level. In the end, everyone is shouting that AI will change the world, but what is really being changed is only click rate and worry. The tech is not wrong. The problem is that humans always like to treat the fastest place to make money as “the most important direction.” If the ad world makes money first, that does not mean this is the most decent use of AI. It only shows that the business world is always more sensitive to “cutting and taking” than to “building.”

I now believe more and more that the more serious a person is about work, the more that person should take control of AI. Because the truly valuable part of AI was never to perform for you. It is to take away the bad work that is not worth your life. It should do first sorting, first counting, first drafts, first grouping, process pushing, repeated checking, and structure sorting. Then people can take their time back and use it to judge, design, create, talk, deal, and decide. AI should be a strong helper. It should not be raised into an electronic god. The person who treats it like a tool will grow stronger and stronger. The person who treats it like a brain will, sooner or later, let his own brain go weak.

Of course 6,000 ads in one day can make money. And yes, it is scary enough. But the louder this kind of “success” gets, the more careful we should be. What it reminds us of is not how great AI is. It reminds us how easily humans take great things and first use them for the shortest-sighted work. If we truly want AI to have some dignity, then first do not rush to use it to fool other people better. First use it to help you do your work in a more solid way, free your mind, and do the part that should really be done by humans in a more human way.