AI Snapshot: ChatGPT and Implications for Workers

Chat Generative Pre-trained Transformer (ChatGPT) is an artificial intelligence (AI) large language model (LLM) released by the company, OpenAI, last year. ChatGPT is “trained” on existing material from the Internet, books and the like, and is quickly able to generate articulate (“human-like”) responses from user prompts. Microsoft recently announced a one-billion dollar investment in OpenAI. Google and other LLM creators are entering the field as well. With the introduction of ChatGPT, there has been speculation about how its proliferation may affect white collar and knowledge workers.

ChatGPT is a disruptive technology; that is, “technology that changes a workflow, disrupting old ways of doing things and generating new ones.” Some suggest that with advancements in AI, machine learning, advanced robotics, and biotech, we have entered into the Fourth Industrial Revolution. Given this, the extent to which ChatGPT (and similar LLM) will complement or compete with white collar and knowledge workers remains to be determined.

A 2019 report from The Brookings Institute, indicates most industries will be impacted by AI. However, it also suggests workers with “graduate or professional degrees will be almost four times as frequently exposed to AI as workers with a high school degree.”  A 2020 CommericalCafe study of U.S. white collar jobs and compatibility with automation, found that “60% of white collar jobs were less than 50 percent compatible with automation.” Also, a 2021 report by the World Economic Forum, suggests that 85 million jobs across industries, will be lost to AI-related change and innovation by 2025. However, the report also predicts some 97 million new jobs will be created within the same timeframe.

The wider impacts of ChatGPT and similar technology remain to be seen. Two things are clear, however. First, LLMs have significant liabilities that may affect proliferation in the workforce. Second, AI will impact specified work tasks and not necessarily jobs or work roles in their entirety.

Host of the Unsung Science Podcast, David Pogue, explored some important questions about ChatGPT (and similar AI) in a recent episode . First, while AI can readily write essays, poems, and the like, podcast guest Professor June Rosenzweig of Harvard University, asks, “Where do we figure out what we think?” Second, ChatGPT is unable to distinguish fact from fiction as it is trained on information from the Internet, which may or may not be factually based. How do we deal with what Stanford University Professor Erik Brynjolfsson, calls “misinformation factories?” Ethical questions abound as well about considerations such as bias in the generated content. Third, there is the notion of “automated plagiarism,” with ChatGPT derived responses being generated from existing information and some creative works written by real people. Finally, are we somehow mistaking writing for creativity since LLMs cannot really think?

How do we deal with these realities, especially when ChatGPT garnered more than one million users shortly after being released to the public? OpenAI also recently announced a commercial version of ChatGPT, ChatGPT Plus, which will be available for $20 per month. Use of ChatGPT is increasing steadily as are concerns about misuse. The company developed a “decoder,” which, with some limitations, is able to detect text written by ChatGPT. Princeton University student, Edward Tian, created GPTZero, which is also able to detect AI generated text. Similar detectors exist as well and OpenAI’s Sharing and Publication Policies, clearly indicate that users be transparent in disclosing any use of AI. Is this enough?

Erik Brynjolfsson, the Jerry Yang and Akiko Yamazaki Professor and Senior Fellow at the Stanford Institute for Human-Centered AI (HAI), likens the use of ChatGPT to a calculator—the calculator for writing. He states that ChatGPT will “augment our ability to write.” He also proffers that ChatGPT may help eliminate rote work and offer workers opportunities to be more creative. This brings us back to the notion of job tasks. A report from the Brookings Institute outlines the following, “A job is a collection of tasks, and even under the most aggressive scenarios, it is unlikely that machines will substitute for all tasks in any one occupation.”

At the “AI and White Collar Jobs” panel at the 2023 World Economic Forum Annual Meeting in Davos, experts agreed that white collar jobs are changing, not necessarily being lost. Panelist Erik Brynjolfsson intimated we are in a period of “the great reconstruction of work,” where the emphasis needs to center on “keeping humans in the loop” and enhancing job quality. He also stated that about one billion knowledge workers exist globally, and anywhere from 15% to 70% of computer work has the potential to be automated.  In addition, Brynjolfsson suggested what is necessary is to determine where humans have a comparative advantage (relative to AI) with respect to certain job tasks such that the technology compliments human work products versus replacing them.

The Davos panelists were in agreement that “reskilling” is critical. According to a recent Forbes article, reskilling is, “providing training to help an employee move into a different job or pursue a different career path.” The counterpart to reskilling is upskilling, which is “providing education and training so employees can improve existing skills and learn advanced skills that help them do their job better or more efficiently.” According to Brynjolfsson, “human capital is a $200 trillion asset in the United States,” and it is important to have employees maximize their potential. Fellow Davos panelist, Mihir Shukla, founder of Automation Anywhere, suggests that it is short-sighted to think that employees can only perform the jobs they are currently occupying.

The substantive impact of AI on white-collar and knowledge workers is yet to be determined. New technologies can both create and eliminate jobs. Reskilling and upskilling are important mechanisms for preparing workers to complete higher-order tasks and use a greater level of their skills and creativity at work. Even with the introduction of ChatGPT4 later this year, which, David Pogue advises, will be trained on 500 times the information as the current ChatGPT3, it remains unlikely that AI will supplant human thought, creative and emotional intelligence.

At The Write Touch, we bring these human dimensions to bear in all of our professional work. We value our clients’ individuality and make sure we incorporate your thoughts, personality, and unique style into the products we develop for you. Contact us today, and let us provide the “human touch” to all of your writing and business needs.

By The Write Touch LLC Writer: Michelle Bragg