本文的封面图也是AI生成的，小编已经失业了（不是
ChatGPT横空出世才几个月，多个行业已经切实感到了来自AI的冲击。
社交媒体上，打工人们纷纷担心自己被裁员。传言中受AI冲击最大的包括游戏美术、插画、文案、服装设计等行业。
我们经过访谈发现，游戏、设计等行业的原画师这个工种受到的冲击最快、最大，已经出现了裁员现象。广州一家中型外包游戏公司的特效技术总监丹青称，公司原画团队2023年初使用AI之后，已经裁员2/3。
还有很多人在考虑转行，原画师小毛转型成为美妆博主，学习动画的大四学生龙三丰则考虑毕业后去当教师。
更多行业暂时还没有受到裁员的直接冲击，但是已经开始运用AI工作。从业十多年的手游项目主策划张立东表示，AI已经能辅助广告公司的前期工作，节省了很多成本。
很多受访对象告诉我们，用AI帮助自己工作也并不是那么容易的事。今后，自身水平过硬，又善于使用AI的“复合型人才”将会更受企业的欢迎。
一、原画师已开始被裁
在丹青所在的公司，原画团队自从2023年初引入国外的一款付费AI软件后，工作效率提升了3倍。借助AI，打草稿的步骤节省下来，原画师只需完成后期细节修改，“相当于节约了50%的工作量。”之前15个人的工作量，现在5个人可以完成。
于是，公司开始分批裁人：每次裁三四个，一共裁了22位，团队人数缩水2/3。公司明确告知员工，裁他们就是因为有了AI。
裁员后，公司每个月能省下20万元人力成本，“现在还只是初级阶段，以后有成熟的制作链就会节省更多成本。”丹青说。有了AI后，公司的订单虽然没有增加，但利润率明显提升。
不过，丹青所在的公司裁的都是初级和中级原画师——他们的工作已经可以被AI完全取代。技术更好的资深原画师则无人被裁。
丹青解释说，资深原画师的核心竞争力在于通过后期调整，来满足甲方的细化需求。“这是AI还没有办法完成的，还需要很长时间的迭代才有可能实现。”。
从事服装设计的悦悦就发现，公司放弃了一个各方面都很合适，只是不会用AI的原画师，并招来三个会用AI做图的画师之后，反而出现了一些新的问题。
首先，团队的整体效率并未因AI而提升太多。使用AI辅助绘画后，前期工作加速了，能省下1/5的时间，但是由于这几位“AI画师”本身水平一般，又要花很多时间去做后期修改。
另一个问题是，用AI可能会被客户发现。
为了维持订单价格，乙方通常不会主动告知甲方自己使用AI作图。但是，自从悦悦的公司用上AI以后，甲方经常发现画稿中有“非人类所能造成”的细节错误，进而发现悦悦的公司使用AI，要求降价。双方开了好几次会也没谈拢价格。
后来，公司把三个AI画师辞退到只剩一个，并要求他优化流程。公司也不招新人了，只敢招一些技术很熟练的老员工。
二、AI很难完全取代人
目前，更多行业和工种里，AI暂时还构不成对人的威胁。
在乙方美术团队内部，原画之外的岗位受AI影响并不大。丹青表示，目前AI还没有波及动作、模型和特效这三个团队。
特效资源库是每个公司的独立资产，在市面上不流通，而AI需要大量资源库喂养。因此，特效行业在很长的一段时间内都不会受到AI冲击。
游戏领域的甲方公司也不太会受到影响。腾讯某头部游戏工作室的一位策划人员表示，像他们这种稳定上线的项目，整体情况还不错，所以基本没有出现裁员的情况，更毋论因为AI的应用而裁员。
游戏制作人张立东表示，现在AI工具所产出的内容，还无法完全满足作品产出的需要。在游戏开发中，需要把人物角色、道具等在Photoshop（专业修图软件）中分层处理，但AI交出来的内容，是单一的图层，所以应用局限很大。
Stable Diffusion和Midjourney是目前常用的两款AI绘图工具。张立东认为，Midjourney是傻瓜式操作，适合普通人；Stable Diffusion则更适合用于游戏行业的图像生产，且功能和可拓展性每天都在进步。但是，要想让它成为彻底的生产工具，“现在还非常困难”。
不仅如此，用AI并不是一件容易的事情。包括GhatGPT、Midjourney等在内的一系列AI工具，都有比较大的使用门槛。“能好好使用AI工具（对人员的要求），可能比某些小公司的用人需求还要高。”
最初使用AI绘画工具时，很多人会讶异于自己通过简单的操作就能获得一张高质量的图，可一旦需求越来越细，难度就会指数级提升，“给AI提需求是一件非常困难，或者说非常有技术性的事情。”
从现有的AI工具及人员水平来看，应用到工作中肯定是会付出比较大的经济代价。“大家需要很多培训学习及试错的成本。”张立东表示。AI画图工具本身是个以算法为基础的大模型，如果想要大规模应用在实际工作中，还需要团队不断编写插件。
在绘图之外，游戏行业的文案工作也没有被AI取代。在某游戏大厂担任文案策划的小鲨表示，公司已经开始研究AI，但她身边也并没有人因此被裁员，“商业写作涉及的内容很复杂，而且游戏文案不是纯写作，涉及策划内容和多部门的配合，这些AI都实现不了。”
张立东告诉我们，AI在文案工作上确实有超高的效率，但它的成文容易被分辨出来。“一般实际工作中所需要的文本，都是特定的，针对某个主题，需要有比较深入或者新颖的观点，目前AI还不太容易做到。”
很多情况下，AI或许可以拿出99％近似对的答案，但也可能会出错，只是它没办法自己检查，永远不知道自己是错的。从这个方面来看，人永远不会被完全取代。“低级的工种可能会被取代，但每一个新技术出来的时候大概都是这样。”张立东说到。
很多公司考虑到版权风险，不会将ChatGPT投入使用。“公司目前还不会将ChatGPT作为辅助编程的工具。”某科技大厂的高级工程师程建也说。一方面，这是因为公司有信息安全、商业机密的规定，不会将软件需求上传至ChatGPT的网络，另一方面，大公司的代码要查重，不能直接引用外部代码片段，如果直接拷贝ChatGPT的生成结果，会存在版权风险。
三、更多行业，AI放大人的能力
“AI取代人不是一件特别容易的事情，但是利用AI工具进行效率提升是很容易的，现在就能实现了。”张立东认为。
包括大厂在内的很多公司已经开始运用AI工作，放大人的能力。
以他刚刚为之做过Stable diffusion使用培训的广告公司为例，该公司已经有意识地将Midjourney等AI绘图工具用在业务的前期沟通阶段，比如先用AI出几张示意图，看看哪个更靠近甲方的需求，就能在这个基础上继续向前推进。“只这一个步骤，就可以为他们节省很多时间。”
阿里一位推荐算法相关业务的主管告诉我们，近一个多月来，他们团队已经经常在工作中使用ChatGPT了，用来帮助写代码和替代搜索引擎的功能。
上述腾讯某头部游戏工作室的策划人也提到，AI的应用会有助于策划或者运营岗的同事更好地跟美术团队进行沟通。“有些需求，语音或者文字可能表达不准确，但是现在就可以用Midjournry，先跑一批符合自己需求的参考图，然后再去跟美术或者UI（界面设计）同事沟通。”
只是现阶段的这种应用，更多是员工的个人行为，公司层面并未倡导。
除了离AI技术很近的游戏和互联网公司，传统制造业也开始大范围运用AI。
或许在普遍认知中，AI与传统制造业相隔甚远，后者的低利润运营模式撑不起前者的高成本。但是，在莆田一众制鞋厂商中间，索罗芬（Soulsfeng）的创始人黄逢春就将AI带进了传统制鞋业。
索罗芬对AI的应用主要在设计和广告环节。设计领域，索罗芬借助Stable Diffusion这个模型从文本生成图像。黄逢春表示，从前至少需要五名设计人员完成的工作，现在只需一名设计师。随着AI的不断“学习”，它对指令的理解和风格的把握也更加纯熟。
如果说产品设计仍需设计师把控风格，那广告营销对AI的倚仗就更多了。黄逢春举了个例子：老板不可能要求一个员工做1000条广告，并控制所有广告的投放渠道、产出效率等，但AI可以，只要设置好投入产出比等几个关键数据即可，AI可以做出几千甚至上万条广告，并分发在合适的渠道。“换做人力，至少也得是几十个人的团队共同协作。”
四、打工人的焦虑和转变
在AI投入应用后，不少打工人陷入焦虑，开始考虑转行，尤其是在原画师这样最先受到冲击的领域。
“AI带来的是精神上的打压，让中低层画师产生自我怀疑。”有四年从业经历的插画师M君说。
之前，M君每个月能画四张图，这个月只画了两张，大部分时间都在焦虑和尝试AI中度过。当她看到自己用AI生出来的图很粗糙，不能用的时候，就可以安心地去画画；但是，在社交平台上看到别人用AI生出非常精美的图后，她就又开始焦虑。
插画博主龙三丰认为，AI对画师的冲击就像方便面对餐饮行业的冲击一样，有人会选择就此吃快餐、走捷径，但她还是会踏踏实实烹饪、练习技术。但是，她在看到关于裁员的讨论时也无可避免地“感到很慌张。”
北京某科技公司程序员小杨发现，很多使用过ChatGPT的前端程序员也都开始慌了：“感觉技术升级后，本来10个人做的工作，现在一个人加ChatGPT就能做。”
小毛是一名工作两年多的原画师，她认为AI会让打工人更加辛苦。“在游戏公司工作本身就很累，每天我要上11个半小时班，AI出现又会进一步加剧行业内卷。”
在精神内耗过后，很多人开始谋求转型。丹青称，特效行业之前人才没有那么饱和，但是因为AI的使用，很多人开始从原画等行业涌入特效行业。他近一个星期收到的简历，有1/5是从原画行业转来的。
龙三丰考了教师资格证，“如果无法找到心仪的美术工作，我还能去当教师。”
去年底被公司劝退后，小毛决定彻底离开游戏行业，做起了自由感更强的美妆博主。
但对于有些人来说，AI带来的焦虑不过是九牛一毛。小杨告诉我们，35岁门槛、互联网行业不景气、疯狂内卷等重重焦虑包围着底层程序员，“很少有人在意再加一点ChatGPT带来的焦虑。”
M君也选择留在插画行业，发挥人的创作在情感上的优势。她使用Stable Diffusion工具，把自己的半成品喂给AI，发现AI没有她自己画的脸有情绪和特色。“虽然我画得不完美，能看出笨拙的痕迹，但我感觉我自己画出的更有情感。”
一家自动驾驶公司的程序员李立也对AI持积极态度。李立认为，对于程序员来说，如何看待AI的影响取决于对自身工作价值的定位。如果程序员认为自己吃的是资源，那ChatGPT毫无疑问威胁了他们的技术资源；如果程序员认为自己的价值来自服务，来自提高效率带来的利润。那ChatGPT作为一个帮助程序员提效的工具，应该受到欢迎。
“一个好的技术不应该是让人失业的技术，而是解放人的技术。”
翻译：
ChatGPT has only been around for a few months, and several industries are already feeling the impact of AI.
On social media, workers are worried about being laid off. Rumors of the biggest impact of AI include game art, illustration, copywriting, clothing design and other industries.
Through interviews, we found that original artists in games, design and other industries have been hit the fastest and the biggest, and there have been layoffs. Dan Qing, special effects technical director of a medium-sized outsourcing game company in Guangzhou, said that the company’s original art team has been laid off by two-thirds after using AI in early 2023.
More industries have not yet been directly affected by layoffs, but are already using AI to work. Zhang Lidong, the main planner of the mobile game project who has been engaged for more than ten years, said that AI has been able to assist the early work of advertising companies, saving a lot of costs.
Many interviewees told us that using AI to help them work is not so easy. In the future, the “compound talents” who are excellent in their own level and good at using AI will be more popular with enterprises.
First, the original artist has begun to be cut
In Danqing’s company, the original painting team has increased its work efficiency three times since the introduction of a paid AI software from abroad in early 2023. With the help of AI, the steps of drafting are saved, and the original artist only needs to complete the details in the later stage, “equivalent to saving 50% of the work.” Five people can do what 15 people used to do.
So the company began to cut people in batches: three or four at a time, a total of 22 people were cut, and the number of team members shrank by two-thirds. The company explicitly told employees that they were being laid off because of AI.
After the layoffs, the company can save 200,000 yuan per month labor costs, “now is just the initial stage, there will be mature production chain will save more costs.” “Said Danqing. With AI, although the company’s orders did not increase, the profit margin was significantly improved.
However, Danqing’s company is only laying off junior and mid-level original artists – their jobs can already be completely replaced by AI. None of the more skilled senior artists were cut.
Danqing explained that the core competitiveness of senior original artists is to meet the needs of Party A’s refinement through later adjustment. “That’s something AI hasn’t been able to do yet, and it’s going to take a long iteration to make it possible.” .
Yue Yue, who is engaged in clothing design, found that after the company gave up an original artist who was suitable in all aspects, but would not use AI, and recruited three artists who would use AI to make pictures, some new problems appeared.
First, the overall efficiency of the team has not been improved much by AI. After using AI to assist painting, the preliminary work is accelerated, which can save 1/5 of the time, but because these “AI painters” are of average level, they have to spend a lot of time to make later modifications.
Another problem is that using AI could be discovered by customers.
In order to maintain the order price, Party B usually does not take the initiative to inform Party A that it uses AI for drawing. However, since Yueyue’s company used AI, Party A often found that there were “non-human” details in the drawings, and then found that Yueyue’s company used AI and asked for a price reduction. The two sides held several meetings but failed to agree on a price.
Later, the company fired three AI artists down to one and asked him to optimize the process. The company does not recruit new people, only dare to recruit some skilled old employees.
Second, it is difficult for AI to completely replace people
At present, in more industries and jobs, AI is not a threat to people for the time being.
Within the art team of Party B, positions other than the original painting are not greatly affected by AI. Danqing said that at present, AI has not affected the three teams of action, model and special effects.
The special effects library is an independent asset of each company, which is not circulated in the market, and AI needs a large number of resource libraries to feed. Therefore, the special effects industry will not be affected by AI for a long time.
Party A companies in the gaming sector are also less likely to be affected. A planner of a Tencent head game studio said that the overall situation of projects like their stable online is good, so there are basically no layoffs, let alone layoffs because of the application of AI.
Zhang Lidong, the game’s producer, said that the content produced by AI tools can not fully meet the needs of the production of works. In game development, characters, props, etc. need to be layered in Photoshop (professional retouching software), but the content handed out by AI is a single layer, so the application is very limited.
Stable Diffusion and Midjourney are two commonly used AI drawing tools.
Zhang Lidong believes that Midjourney is a fool’s operation, suitable for ordinary people; Stable Diffusion is better suited for graphics production in the gaming industry, and features and scalability are improving daily. But to make it a full-fledged production tool, “it’s very difficult right now.”
Stable Diffusion page image from the network
Not only that, using AI is not an easy task. A series of AI tools, including GhatGPT, Midjourney, etc., have relatively large thresholds for use. “Being able to use AI tools well may [require people] more than some small companies need.”
When initially using AI painting tools, many people will be surprised that they can get a high-quality drawing through simple operations, but once the demand is more and more detailed, the difficulty will be exponentially increased, “it is a very difficult or very technical thing to give AI needs.”
From the existing AI tools and the level of personnel, the application of AI to the work will certainly pay a relatively large economic price. “There’s a lot of training and learning and trial and error costs.” Zhang Lidong said. The AI drawing tool itself is a large model based on algorithms, and if you want to apply it on a large scale in actual work, the team needs to constantly write plug-ins.
Outside of graphics, copywriting in the game industry has not been replaced by AI.
In a big game factory as a copywriter planning small shark said that the company has begun to study AI, but no one around her has been laid off, “the content of business writing is very complex, and the game copy is not pure writing, involving planning content and multi-departmental cooperation, these AI can not be achieved.”
Zhang Lidong told us that AI does have ultra-high efficiency in copywriting, but its writing is easy to distinguish. “The text required in general practical work is specific, and it needs to have a relatively in-depth or novel viewpoint for a certain topic, which is not easy for AI to do at present.”
In many cases, AI may be able to come up with 99% approximate correct answers, but it may also be wrong, but it has no way to check itself, never know that it is wrong. In this respect, people can never be completely replaced. “Low-level jobs may be displaced, but that’s probably the case with every new technology that comes out.” Zhang Lidong said.
Many companies will not put ChatGPT into use due to copyright risks. “The company is not using ChatGPT as a programming aid at this time.” Cheng Jian, a senior engineer of a large technology factory, also said. On the one hand, this is because the company has information security and trade secret regulations, and will not upload software requirements to the network of ChatGPT. On the other hand, the code of large companies needs to be checked and can not directly reference external code snippets. If you directly copy the generated results of ChatGPT, there will be copyright risks.
Third, more industries, AI magnifies people’s ability
“It is not particularly easy for AI to replace people, but it is easy to use AI tools to improve efficiency, and it can be achieved now.” Zhang Lidong thinks.
Many companies, including big factories, are already working with AI to amplify human capabilities.
Take the advertising company for which he has just done training on the use of Stable diffusion as an example, the company has consciously used AI drawing tools such as Midjourney in the early communication stage of the business, such as first using AI to produce several sketches, to see which is closer to the needs of the party, and then continue to move forward on this basis. “That one step alone can save them a lot of time.”
A director of Alibaba’s recommendation algorithm related business told us that for more than a month, their team has often used ChatGPT in their work to help write code and replace search engine functions.
The above-mentioned planner of a Tencent head game studio also mentioned that the application of AI will help colleagues in planning or operations to better communicate with the art team. “There are some requirements that may not be accurately expressed in voice or text, but you can use Midjournry now, run a set of reference images that meet your needs, and then communicate with your art or UI colleagues.”
However, this application at this stage is more of an individual behavior of employees and is not advocated at the company level.
In addition to games and Internet companies that are close to AI technology, traditional manufacturing industries have also begun to use AI on a large scale.
Perhaps in the general perception, AI is far from traditional manufacturing, and the latter’s low-profit operating model cannot afford the former’s high cost. However, among the Putian shoe manufacturers, Huang Fengchun, the founder of Soulsfeng, brought AI to the traditional shoe industry.
Soprofen’s use of AI is mainly in design and advertising. In the field of design, Solofin uses the model of Stable Diffusion to generate images from text. What used to take at least five designers now takes just one, says Mr Huang. As the AI continues to “learn,” its understanding of instructions and style also become more sophisticated.
If product design still needs designers to control style, advertising marketing relies on AI even more. Huang Fengchun gave an example: the boss can not ask an employee to do 1,000 advertisements, and control all advertising channels, output efficiency, etc., but AI can, as long as the input-output ratio and other key data can be set. AI can make thousands or even tens of thousands of advertisements, and distribute them in the appropriate channels. “In terms of manpower, at least dozens of people have to work together.”
Fourth, workers’ anxiety and change
After AI was put into application, many workers became anxious and began to consider changing careers. Especially in the field of original artists, which was the first to be affected.
“What AI brings is mental repression, which creates self-doubt among middle – and lower-level artists.” Said M Jun, an illustrator with four years of experience.
Before, M Jun could draw four pictures a month, but this month only drew two. And most of the time was spent in anxiety and trying AI. When she saw that the drawings produced by her AI were very rough and could not be used, she could draw with peace of mind. However, after seeing others using AI to produce very beautiful pictures on social platforms, she began to worry again.
Illustration blogger Long Sanfeng believes that the impact of AI on artists is like the impact of convenience on the catering industry. Some people will choose to eat fast food and take shortcuts, but she will still cook and practice technology. But she also, inevitably, “gets flustered” when she sees talk of layoffs.
Xiao Yang, a programmer at a Beijing technology company, found that many front-end programmers who had used ChatGPT also began to panic: “I feel that after the technology upgrade, the work that was originally done by 10 people can now be done by one person with ChatGPT.”
Xiao Yang used Chat GPT to solve some work problems
“Working in a game company is tiring in itself. I work 11 and a half hours a day, and the advent of AI will further aggravate the industry.”
In the wake of spiritual inactivity, many people begin to seek transformation.
Danqing said that the special effects industry was not so saturated with talent before, but because of the use of AI. Many people began to pour into the special effects industry from original painting and other industries. One fifth of the resumes he receives in the past week are from the original painting industry.
“If I can’t find an art job I want, I can still be a teacher,” said Long, who got a teaching certificate.
But for some, the anxiety caused by AI is just a drop in the bucket. Xiao Yang told us that the 35-year-old threshold, the depression of the Internet industry, crazy internal volume and other anxieties surrounded the underlying programmers, “few people care about adding a little bit of ChatGPT anxiety.”
M Jun also chose to stay in the illustration industry to give play to the emotional advantages of human creation. Using the Stable Diffusion tool, she fed her semi-finished product to the AI and found that the AI did not have the mood and character of the face she had drawn herself. “Although I don’t draw it perfectly and can see signs of clumsiness. I feel that my own paintings have more emotion.”
Li Li, a programmer at an autonomous driving company, is also positive about AI. Li Li believes that for programmers, how to view the impact of AI depends on the value of their own work. If programmers think they’re eating resources, ChatGPT certainly threatens their technical resources. If programmers think their value comes from the service, from the profit that comes from increasing efficiency. ChatGPT should be welcomed as a tool to help programmers improve performance.
“A good technology should not be a technology that puts people out of work, but a technology that liberates people.”
本文由数字化转型网（www.szhzxw.cn）转载而成，来源于财经杂志；编辑/翻译：数字化转型网宁檬树。
免责声明: 本网站(http://www.szhzxw.cn/)内容主要来自原创、合作媒体供稿和第三方投稿，凡在本网站出现的信息，均仅供参考。本网站将尽力确保所提供信息的准确性及可靠性，但不保证有关资料的准确性及可靠性，读者在使用前请进一步核实，并对任何自主决定的行为负责。本网站对有关资料所引致的错误、不确或遗漏，概不负任何法律责任。
本网站刊载的所有内容(包括但不仅限文字、图片、LOGO、音频、视频、软件、程序等) 版权归原作者所有。任何单位或个人认为本网站中的内容可能涉嫌侵犯其知识产权或存在不实内容时，请及时通知本站，予以删除。
