Will AI office assistant reduce the burden on workers or add to it?

With the development of AI, some people have regarded AI as an important part of production, like finding an emotionally stable assistant for themselves, but AI assistants also have limitations.

Zhang Yi is a doctoral student in physics in Beijing. One day, he needed to convert the language MATLAB to Python.

This is an auxiliary work for scientific research. Writing code is not his specialty. It usually takes two hours for him to do this. His usual practice is to change the syntax line by line according to the code. But he thought of ChatGPT, maybe he could give it a try.

In the dialog box, type "I will give you a piece of Matlab code, and you help me change it into Python." After a few seconds, the code was completed. Zhang Yi was very surprised. Although it was not completely accurate, it only needed to be manually verified, which did not take too much time. After using it for nearly a year, Zhang Yi exclaimed to Xinmei Daybreak, "It saved 80% of the time in the past."

ChatGPT has ignited a wave of AI production transformation. This should not be a one-man show by entrepreneurs, but also requires the witness of users.

While some people are still worried about AI replacing humans, some people have already regarded AI as an important part of production, just like finding themselves a well-trained and emotionally stable assistant for free.

Students use AI to write codes and papers; judges use AI to write documents; designers use AI to make pictures and even make money. After AI completes its work, they remember to say thank you to AI.

Although they also acknowledge the limitations of AI, AI is currently more suitable for handling mechanical work. In addition, whether you can get along with AI happily depends on whether you use the prompt words well.

But we have to admit that the impact of AI on the production side may have just begun.

1. An emotionally stable work assistant

Zhang Yi now uses ChatGPT every day. "I can't live without it."

For Zhang Yi, AI has become a powerful assistant to improve his work efficiency. Daily research often requires writing code to assist work. But because Zhang Yi is not a computer major, some codes are very difficult to write. This takes up a lot of research time.

Last year, ChatGPT sparked AI fever, and a friend suggested that Zhang Yi give it a try. Miraculously, Zhang Yi found that AI was much easier to use than he had imagined. Some classic codes, such as sorting algorithms, can be fully handled by AI, and there are basically no mistakes. The verification method is also very simple, just find a special value to calculate.

Basic physics knowledge can also be asked directly to AI. In addition, Zhang Yi told Xinmei Daybreak, "You can also change your poor Chinese English into fluent English."

Although it cannot make a direct contribution to scientific research results, AI still saves Zhang Yi a lot of mechanical work, so that he can free up more time to concentrate on thinking. Zhang Yi found that many of his fellow students have begun to use AI, some using ChatGPT and some using Wenxin Yiyan.

As a handy tool, AI has a wider range of applications.

Guo Xiao is a judge's assistant. Her daily work is to write judgments and do research. In addition to work, they also receive other tasks, such as writing a reflection on a current political topic.

In the past, her routine was to use search engines to collect information and then write an article based on her own ideas. After using AI, a logically clear manuscript was directly generated. Guo Xiao only needed to read the entire article and add her own opinions.

Usually, this type of work is a marginal task, and the quality will not affect work performance. Using AI is efficient and there is no need to worry about making mistakes. Guo Xiao expressed it more directly to Xinmei Daybreak, saying that its role is to complete "work that is not very useful but must be done." However, this type of task usually only occurs once a month.

In addition to code and text, the potential of AI in the field of audio and video is also being explored and verified. At the beginning of the year, Jianying updated the new function "clone voice". Users can generate their own exclusive audio by reading the provided example sentences according to the instructions. Many people said, "I don't have to read the lines myself anymore." Repeated recording and inaccurate pronunciation are the pain points of many bloggers.

AI digital people have also been applied in the commercial field. On April 16, Liu Qiangdong's AI digital person made his first live broadcast on the JD App. Within 20 minutes of the broadcast, the number of viewers in the live broadcast room exceeded 9 million. This is not only a sales event, but also a marketing event. More businesses will realize the benefits of AI digital people: realistic appearance, smooth interaction and low cost.

In addition to content generation, code generation, image generation, and virtual human interaction, the application scope of generative AI is gradually expanding to game development, advertising marketing, and even industrial design, penetrating into all walks of life. Many people sigh, "AI is really awesome now." "The future is here."

Zhu Jun, vice president of ByteDance, also expressed confidence when sharing Doubao's product design, "On the one hand, the big model technology itself is constantly improving, and on the other hand, the industry's exploration of application forms and interaction methods will gradually mature, so that big model products will gradually be integrated into the lives and work of more users."

In corners that most people have not yet seen, the influence of AI is spreading. Some people have absorbed the nutrients first, and some have already smelled business opportunities.

2. Is it difficult to monetize through AI?

Nana's motivation for getting involved in AI was interest and curiosity, but she also received monetary rewards while creating content.

When Nana first opened the AI ​​painting software Midjourney, she didn’t define it too much. She simply typed in five words, “rabbit in the wind.” When the image came out, her first feeling was amazing. “I felt the creativity of AI.”

This started Nana's AI exploration journey. She opened an account on Xiaohongshu called "A Little Nana" and focused on AI photography.

The operation principle of Wenshengtu is simple, but in order to make the generated image more exquisite and creative, it is necessary to combine photography knowledge with skillful prompt words. For example, the prompt word for Nana's picture is, "Spring, early morning, transparent light, a female model in a yellow dress sitting by the lake, the trees are full of flowers. Female model with charming features, half-length photo, taken by timwalker."

Nana's main job as a photographer quickly opened the door to AI painting. However, after three or four months of continuous posting, there was no significant increase in traffic. Fans seemed to be more interested in her fashion show and clothing design pictures. She decided to change direction and try to expand AI clothing.

Although both are AI painting, AI clothing and AI photography have different weights. AI photography emphasizes light, environment, emotion and atmosphere, while AI clothing focuses more on design style, design elements, style matching, tailoring, materials and other more detailed and concrete contents. This requires not only a basic knowledge of photography and modeling, but also an understanding of the style characteristics of different periods and the historical culture behind different styles, as well as a more divergent thinking and aesthetic collision.

AI Clothing

Because she has been interested in fashion design since she was a child and has worked with fashion stylists at work, Nana has invested great enthusiasm in learning, experimenting and iterating at the same time. "AI's fast expression has also made me more self-motivated to learn." Sometimes she would work for three or four hours until she was satisfied.

When her fans reached 7,000, business came to her. She told Xinmei Daybreak, "I couldn't believe it." She was excited to get feedback after working hard for half a year, but also felt the pressure of working with a jewelry brand worth tens or even hundreds of millions. After receiving the brief, Nana concentrated on learning design and typesetting, not daring to slack off.

More and more brands are looking for cooperation, and the monetization methods are more diverse. Some are easier than brand promotion, and do not require much communication, and can directly sell the copyright of the pictures. But in general, the sum of various incomes is not as good as the main business, and it is not stable. But for Nana, she values ​​the interest and the sense of accomplishment brought by the extra income more.

Nana is not an isolated case. The commercial potential of AI has generated a variety of monetization channels.

Blogger Midstudio focuses on AI furniture design on Xiaohongshu, breaking the traditional design rules. AI furniture design allows for the release of imagination and creativity. For example, the blogger incorporates fruit elements into the design of a chair, which weakens the image of the chair and looks more like a cute fruit figurine. Fans praised it in the comment section and asked "where to buy it".

This has become a monetization channel for bloggers: they can work with factories to sell physical goods directly in Xiaohongshu stores. In addition, the accumulated traffic has also brought bloggers many opportunities to cooperate with brands.

More people have discovered the opportunity to monetize AI. From small things like red envelope covers and customized mobile phone cases to large things like art micro-spray prints and customized gift boxes, as long as there are buyers, they can be converted into currency. Some people also use AI to design wedding venues, opening up new business opportunities in the field of wedding planning; others use AI to generate popular science copywriting, accumulate fans, and then promote products...

3. Limitations of AI Production

How big is the application capability of AI? This is a difficult question to answer. Everyone has infinite expectations for AI, but for now, AI still has some thresholds and limitations.

AI can help Zhang Yi save time from tedious mechanical work and improve efficiency, but it cannot provide ideas and directions for his research. Even if he can free up ample time to think, can he speed up the progress of scientific research? Zhang Yi's answer is rigorous, "The output has increased slightly."

Guo Xiao, who used AI to complete her homework, had the same feeling. The advantage of AI writing is that the theme is clear and the structure is clear. But the disadvantage is that it is too general and the viewpoints are not precise enough. Formal legal documents such as judgments, which are Guo Xiao's main work, require strict logic and clear facts. Even the slightest text error may cause irreparable losses.

More importantly, "the final decision of the case must be made by the judge, and the judge's free judgment must be reflected. AI cannot make decisions."

Official document writing is considered one of the areas most likely to receive help from AI. Many AI software have launched official document writing sections, taking on administrative document services such as notices, announcements, and speeches.

Banyuetan quoted a civil servant as saying that AI writing official documents does not fully and accurately understand the basic situation of the local area, and does not understand the local new policies and new proposals, so the generated articles lack eye-catching points. Government materials often require a combination of theoretical knowledge and practical cases, which leads to the limitations of AI writing official documents.

Compared with the text and code fields, AI can only complete basic tasks that are instrumental and objective. In the field of painting, it seems that the imagination can be more abundant. However, the prompt words are still a threshold. The preparation for expression requires a full understanding of the basic common sense of painting, art masters, and the techniques used. This is why the most profitable AI application at present is selling courses.

Not only in the field of painting, the level of use of prompt words in tasks with slightly higher difficulty will affect the final effect of AI presentation. Many people are therefore blocked from the door.

Kevin is an AI enthusiast. He shares his notes on using AI to solve various problems on his Xiaohongshu account "Kevin's AI Practical Notes", such as using AI to "disassemble Xiaohongshu's popular articles with one click" and "generate a 100-point pictorial in 1 minute". In order to make the results more in line with expectations, Kevin replaced general prompts with structured prompts when entering prompts. Some examples are shown below (the content is downloaded from AI software Kimi):

Structured prompt words are believed to improve AI processing efficiency and complete more accurate and even complex tasks that require logical reasoning. Its underlying logic is to publish tasks in a format that uses large models for data training. It can be understood in layman's terms as, I communicate with you in your own words.

But it takes time and effort to understand the concept and usage of structured prompt words. In Kevin's words, the external framework is easy to understand. The key lies in the logic behind the prompt words and whether your expression can be well understood and executed by AI.

In most people's minds, AI means convenience and efficiency. Getting the desired results directly through conversation is their original intention of using AI. Structured prompts undoubtedly lengthen the use path and difficulty of AI. Kevin also frankly said, "It's easy to get started but difficult to master. At least I don't think what I have written is very good at the moment."

AI applications are still waiting for an opportunity for a big explosion.

(At the request of the interviewees, Zhang Yi and Guo Xiao are pseudonyms in this article)