Before the Chinese New Year, an article from Harvard brought reflections.
There is an article at Harvard University in the United States about the possibility of ChatGPT as a substitute for writing work.
The content of the article revolves around a conversation between a hired writer at Harvard University and a well-known psychology professor, and focuses on a basic point:
AI large language models like ChatGPT may help people find information more efficiently or complete searches more creatively. But if people worry that it will replace people to finish writing, this is redundant. Because it can't have a deep intuition about causal connections like people do, not to mention that ChatGPT's answers often don't stand up to logical scrutiny and go against common sense.
Writing often involves expressing emotions, personal opinions, and capturing the complex state of human life. These aspects are deeply ingrained in human consciousness, and it is challenging for large language models to convincingly replicate human capabilities in these aspects.
The psychology professor at Harvard argues that fear of the AI technology stems from assumptions about the worst-case scenario, rather than the possible solutions in reality. For example, in terms of the limitations of large language models, ** institutions or publishers in the United States and other countries have clearly asked journalists and writers to stop using any artificial intelligence applications to write content directly, and many high-tech companies have also developed specialized technologies that can detect whether articles are written by large language models.
It can be said that for the technology of writing, the large language model that will become popular all over the world from 2023 does not seem to be able to replace humans in a short period of time, at most it is just a retrieval tool or search tool, similar to WPS software.
We know that in all walks of life, the main function of writing work is to draft all kinds of documents and documents, write content, and organize or proofread all kinds of articles. This requires a person to write well, be logical, and be able to write articles. Don't underestimate these jobs, an Internet celebrity once boldly exclaimed: "Liberal arts majors are almost worthless." This is actually a very impolite and extreme view, and the skill and judgment required of a good writer are no less difficult than that of a mechanical engineer or a pharmacist.
So, does the above Harvard professor's view mean that current writers do not need to be wary of or think about the threat of large language models or other artificial intelligence applications in the short term?
On the other hand, for artificial intelligence large language models, will there be a breakthrough in the development of 2024 and gradually begin to replace some text work?
Why don't you hear what ChatGPT has to say first
To figure this out, in addition to the opinions of the human experts mentioned above, I thought of asking what AI itself thinks.
I asked a friend from overseas to use ChatGPT to try to answer the above question on my behalf.
My question was straightforward: "Will large language models (LLMs) do the work?" ”
The answer: "As an AI language model, I can help with generating text and assisting with various writing tasks." However, it is unlikely that an AI model like ChatGPT will completely take the author. While AI can help generate content quickly and efficiently, human creativity, insight, and subjective experience are still valuable in many forms of writing.
AI can be an invaluable tool for writers, providing advice, aiding with research, and even generating content drafts. It can save time and provide other perspectives, but it lacks the deep human understanding, emotion, and originality that comes from lived experience.
Writing isn't just about writing words, it's about telling a story, capturing unique images, and connecting with the reader on an emotional level. These aspects often require human intuition, empathy, and creativity, and it is currently difficult for AI models to fully replicate these aspects.
While AI may have the ability to automate certain repetitive or formulaic writing tasks, humans may continue to play an important role in creating engaging, thought-provoking, and nuanced content creation for the foreseeable future. ”
His answer was strikingly consistent with the views of the Harvard professor above, and gave me the following two implications:
1) Writing can be divided into formulaic writing and creative writing from a work perspective. AI large language models can often use large amounts of text data to train their deep learning capabilities so that they can understand the meaning of language text and generate language text.
Therefore, it is possible to write more formulaic textual content (such as simple mechanical documents) instead. But once the writing is complete, it has to be reviewed by a human to avoid mistakes. And the more creative content can't be done by any means. The simplest example is that if the script of the TV series "Flowers" is adapted by ChatGPT based on the original author's name**, it is impossible to succeed. If the reader does not believe it, he can find an opportunity to use a large language model to imitate the poems of Li Bai or Du Fu for creation, and the result will definitely lose all the aura, romance and emotion, and become chicken ribs.
2) On this issue, the views of the Harvard professors are very consistent with ChatGPT's own views, which seem to suggest that some ideas derived from large language models based on massive data training and deep learning may be valuable. After all, it's not often that we have Harvard professors around us to help us clear our minds and provide perspectives, but using large language models is very convenient. This seems to show once again that large language models are real for the ancillary function of writers.
As a result, I think that large language models, while not a complete replacement for writing, do affect some positions that are more formulaic in writing.
or text work that can be replaced by large language models
Last year, there was a piece of news in the United States that caught a lot of attention.
A part-time American employee who has long been engaged in textual descriptions for companies** and products has been laid off by nine companies that provide services to him, on the grounds that these textual descriptions are more mechanical and formulaic, and ChatGPT can be used to do it for him.
The employee later said in an interview with foreign media that one of the companies that laid off his service later came back to him, hoping that he could return to work, because the company was not satisfied with the content produced by ChatGPT, but the other original companies did not look for him again.
Affected by this, he had to change careers, and for this reason, he also suffered some psychological effects, he went online to study the most unlikely to be replaced by chatgpt, and instead engaged in the first job, and is receiving technical training as an air conditioner maintenance worker and plumber. These jobs are said to be at the top of the "safety list".
In early 2023, an analysis by Goldman Sachs pointed out that the generation of AI large language models could seriously disrupt the labor market, affecting about 300 million jobs worldwide. Among them, white-collar jobs in the administrative and legal professions, which are engaged in formulaic and mechanical creations, are considered to be the most risky.
At this point, if you're thinking about working on something mechanical, or if you're working on something that is more formulaic, think about whether the threat of AI large language models is imminent?
What's the solution?
Is there any other way to avoid being replaced, other than changing careers like the clerk in the United States? I have consulted the relevant articles of some famous university literature professors, and summarized the following points for your reference:
1.Overall, many experts believe that AI is not a threat but a tool for writers and writers.
2.Focus on developing our creative and logical abilities. AI can generate text and content, but it cannot replicate the creativity and originality of human writing.
3.Gain insight into your target audience. What kind of content do they want to read? What are their pain points? What kind of language do they like to use? Artificial intelligence often fails to achieve this focus.
4.Strive to be an expert in your field. The more you know about your writing topic, the better equipped you will be to write high-quality content that is informative and engaging. For example, a front-line worker who has worked in the banking system for many years can definitely write more details of bank products or bank strategies when he is transferred to administrative writing.
5.Learn how to use AI-powered writing tools to your advantage. Embrace large language models instead of rejecting them, as new automation technologies can help people automate tedious tasks in the writing process, freeing up their hands to focus on creativity.
6.There is always no substitute for communication between people, and it is necessary to cultivate interaction and communication with various related people. Listen to other people's opinions and insights, and even recommend them directly to the direct readers of the text so that people can recognize the content of their writing.
Disclaimer] The content of the article is for research and study purposes only and does not constitute any investment advice.