"After the boss asked me to work with AI, I turned into a 'guinea pig': the workload doubled, but the salary remained the same"

Source: "CSDN" (ID: CSDNnews), author: Zheng Liyuan

Image source: Generated by Unbounded AI tool

ChatGPT was born at the end of November last year, and the topic of how AI can help humans improve productivity has never stopped, and the evaluation of AI in the scientific and technological circles has also reached unprecedented heights:

▶ In January of this year, Microsoft CEO Satya Nadella said in a conference call: "The wave of AI will affect all aspects of the technology stack, providing people with new solutions and creating new opportunities. Whenever we consider new opportunities for platforms, AI is the answer."

▶ In March, Bill Gates mentioned in his blog GatesNotes: "In the next few years, the main impact of AI on work will be to help people do their jobs more efficiently."

▶ By April, a group of researchers from Carnegie Mellon University, Yale University, and the University of California, Berkeley also found that after testing and comparing, in terms of data set annotation performance, compared with the excellent human annotators they hired, using OpenAI’s GPT-4 saved a total of more than 500,000 US dollars and 20,000 man-hours.

▶ ……

Based on the above remarks and research data, it seems to be an undoubted fact that AI can make people "do work more efficiently". Therefore, many large technology companies are rushing to the AI track, betting on new AI-driven tools that are expected to simplify work.

Executives at companies are clamoring for AI tools, but that may not be the case from the perspective of ordinary employees: Ivana Saula, research director for the International Machinists and Aerospace Workers Union, says workers in her union say they feel like guinea pigs when their bosses rush to roll out AI tools at work.

AI can only do part of it, and the rest makes humans "work harder"

In a recent study, Shakked Noy, a Ph.D. student in MIT's economics department, found that while AI does have important applications at work, "it is too early to say whether it will be good or bad, or how exactly it will cause social change."

As Shakked Noy put it, Ivana Saula pointed out that although AI can replace some aspects of human work, there are still "some unfinished tasks that need to be handed over to human employees to complete." According to Ivana Saula, many employees in his union complained that after adopting new AI tools, their "workload is greater" and "work intensity is heavier" because "now it is all set by the machine."

Specifically, the participation of AI tools in daily work often leads to more "remaining tasks that humans still need to complete", such as additional logistics tasks that machines cannot complete at all, which adds more pressure to humans' daily workflow.

Ivana Saula said that most employees want to be "really involved in the implementation process of AI tools" and hope that bosses can "understand the actual situation on the front line", but the reality is: "A lot of times, the connection between front-line employees and upper management, let alone the CEO, is completely disconnected."

"Overwhelming" AI output, reviewers regard it as "thorn in the eye"

In addition to the fact that AI can only do part of the work, a large amount of the rest can only be done by humans, and the quality of AI output is uneven, which has also brought additional burdens to many industries. The media industry may have the greatest impact.

The generative AI represented by ChatGPT has extreme advantages and disadvantages for media practitioners: the emergence of such AI tools has accelerated the efficiency of copywriting and writing; but the articles generated by AI are also overwhelming, and the huge quantity and quality cannot be guaranteed. Reviewers are very "headaches" - Neil Clarke, editor of the well-known science fiction magazine Clarkesworld Magazine, is one of the "victims".

In February of this year, Neil Clarke announced that due to the recent large number of submissions generated by AI, the magazine decided to temporarily close the online submission channel: "The current submission channel has been closed, and the reason should be easy to guess."

Neil Clarke posted a picture on Twitter, counting the number of authors who have been blocked by Clarkesworld every month due to plagiarism or AI-generated articles in recent years: In January this year, Clarkesworld blocked more than 100 people, and the number of people blacked out exceeded 500 before February was over.

In this regard, Neil Clarke said helplessly: his team has been overwhelmed by a large number of AI-generated submissions.

Because of these submissions, Neil Clarke and his team had to manually review hundreds of pieces of AI-generated content that was generally “terrible” and the quality of the variance wasn’t even the biggest issue, it was more of a “quantity issue”: “It nearly doubled our workload. Over the past few months, we’ve had people talking about efficiency-enhancing AI tools being a thorn in our side.”

For this reason, many people recommend Neil Clarke to use "magic to defeat magic", that is, to use AI tools to identify AI output, but Neil Clarke has not tried it: "These tools are not very useful, because the way they mark whether it is generated by AI is not reliable, and it is especially unfriendly to writers whose native language is not English."

Neil Clarke said with emotion: "You hear those AI experts say that AI will make amazing breakthroughs in different fields, but in fact these are not the fields they are currently engaged in."

Workload doubled, but salary unchanged

On the one hand, AI has caused humans to "double the workload", on the other hand, the wages of ordinary employees who have endured all this have not changed-this is the biggest reason why they think they are "guinea pigs".

Mathias Cormann, secretary-general of the Organization for Economic Co-operation and Development (OECD), recently published a survey report showing that for non-AI experts and non-managers, the impact of using AI on wages is "negligible."

According to the report, employees with AI-specific skills—that is, those who develop, train, or maintain AI systems—earn high salaries and enjoy a substantial wage premium. Among them, the highest wage premium due to AI is in management occupations, which shows that there is currently a large demand for talents who understand how to integrate AI into wider production processes.

But for more ordinary employees who directly use AI tools, the impact of AI on their wages has been almost "negligible" so far. In other words, for most ordinary employees, the use of AI will increase their work intensity, but their wages will not increase accordingly, and even one-fifth of the respondents said that AI reduces their autonomy in work.

Regarding this phenomenon, the OECD stated in the report: "This shows that the impact of AI on productivity has so far been modest. At present, larger and more capital-intensive enterprises (these enterprises are inherently high in productivity) are more likely to adopt AI technology, but after accounting for observable differences between enterprises, the increase in productivity is actually small. "

So, as a programmer, have new AI tools been added to your work in the past six months? Does their appearance improve efficiency or burden for you?

Reference link