Since the emergence of ChatGPT, everyone believes that many jobs in the future will be replaced by AI.

But you definitely can't imagine that it has already happened.

In February this year, a foreign blogger named Henley Wing Chiu, in order to figure out the question of "which jobs are more likely to be replaced by AI," specifically studied the actual freelance job data on Upwork from November 1, 2022 (one month before the release of ChatGPT) to February 14, 2024, and drew the following conclusions:

1. Since the release of ChatGPT, writing, customer service, and translation jobs on Upwork have been the most affected, with translation suffering a double blow in terms of quantity and price: not only has the number of jobs decreased by 19%, but the hourly wage has also dropped by more than 20%.

2. Surprisingly, the demand for jobs such as web design, graphic design, software development, and video production has not declined, but has even increased, with the hourly wages for graphic design and web design jobs also showing a slight increase.

3. The development of AI has also brought many new job opportunities. Contrary to many people's expectations, there has been no increase in data annotation and machine learning jobs, but rather a 2000% surge in the number of jobs related to developing chatbots.

/ 01 / After the release of ChatGPT, the demand for writing, translation, and customer service dropped significantly.

The blogger selected the most popular 12 job categories on Upwork and analyzed the 84-day moving averages of these categories within approximately 3 months. Surprisingly, since the release of ChatGPT, the number of jobs in most categories has actually increased, but the number of jobs in 3 categories has dropped significantly.

The number of new jobs in each category on Upwork since the release of ChatGPT

The three categories with the largest decline in job numbers are writing, translation, and customer service. The number of writing jobs decreased by 33%, translation jobs decreased by 19%, and customer service jobs decreased by 16%.

/ 02 / The number of video editing jobs did not decrease but increased by 39%

Apart from the aforementioned job categories, most other categories have not been negatively impacted—in fact, the number of jobs has increased.

Changes in the number of Upwork jobs since the release of ChatGPT

Since the release of ChatGPT, the number of video editing/production jobs has increased by 39%, graphic design jobs by 8%, and web design jobs by 10%. Software development jobs have also increased, with backend development increasing by 6% and front-end/web development by 4%.

Since I've heard many stories of people using ChatGPT to generate code, illustrations, and even complete videos, I was surprised to see an increase in the number of jobs in graphic design, video editing/production, and even software development. I think these data can be explained from several aspects:

One possibility is that these generative AI tools are already good enough to replace many writing tasks, whether it's writing articles or social media posts. But for other tasks, such as video and image generation, they are not yet perfect. Personally, I needed many prompts to be satisfied with the images created by DALL-E for my blog posts, and that's just because I'm not very picky.

So, if you're creating images that require a lot of detail, these AI tools may not be good enough. Similarly, for videos, we've seen some tools that seem to be able to create many complex video scenes, but can they really create a complete trailer or customer tutorial video? Not yet, although this may change in a few months.

Another explanation is that users need time to learn how to effectively use these tools and understand their abilities and limitations. If we look at the chart of the decline in the number of writing jobs, we will find that it began to decline shortly after the release of ChatGPT, but it wasn't until August that the rate of decline began to accelerate (almost ten months after the release of ChatGPT).

/03/Translation hourly wages have dropped by over 20%, with graphic design and web design being the most resilient

Is there a correlation between a decrease in the number of jobs and a decrease in salaries? Unfortunately, there is no publicly available method to accurately determine the actual salary earned by freelancers, so in this study, I must extract the estimated hourly wage range specified in the job posting, which will be an approximate estimate of the actual income earned by freelancers.

The hourly wage for translation work has decreased by more than 20%, with the most severe impact, followed by video editing/production and market research. Graphic design and web design work are the most resilient. They not only increase in quantity, but also slightly increase in hourly wages.

Overall, I believe it is difficult to draw definitive conclusions from these data. Many salary changes may only be normal seasonal fluctuations. However, a clear conclusion is that there is still a need for graphic design and web design work, which has not yet been replaced by AI tools.

Similarly, I believe this is because tools like DALL-E and MidJourney require some knowledge and creativity. This is evident in some job postings, which explicitly require the search for AI image creation experts. Unlike writing articles, using AI tools to create visually stunning images requires understanding how to construct good prompts, how to adjust prompts, and a lot of experimentation and errors.

/04/Chatbot positions have exploded, with a surge in demand of 2000%.

Okay, we already know which jobs are experiencing a decrease in demand, but due to the development of AI, which jobs are actually experiencing an increase in demand?

My first reaction is that the demand for data annotation work should increase to help train and fine tune Large Language Models (LLMs), and freelance platforms are undoubtedly the best choice for those who find annotated data. Although I don't expect most companies to recruit many machine learning experts on Upwork, the number of machine learning jobs should at least increase, right?

The result proves that both hypotheses are completely wrong. Although data annotation work initially increased, the number of data annotation work has remained stable over the past 10 months. Since the release of ChatGPT, the number of machine learning tasks has actually decreased.

Perhaps the company has recruited more machine learning engineers outside of Upwork. After all, who would hand over important and professional machine learning work to freelancers? However, even if we look at the world's largest companies, the demand for machine learning has not significantly increased.

On the other hand, the work of generating AI content, developing AI agents, integrating OpenAI/ChatGPT APIs, and developing AI applications is becoming increasingly popular. But what is the most popular use case? It's a chatbot. Since the release of ChatGPT and OpenAI APIs, the number of tasks related to developing chatbots has surged by 2000%. If there were a killer use case for AI today, it would be developing chatbots.

I think one conclusion that can be drawn from this is that most companies do not engage in exciting work, such as developing their own Large Language Models (LLMs) or using a large amount of training data for adjustments. On the contrary, they are more focused on integrating OpenAI APIs into existing products and developing chatbots to replace customer service agents.