Abstract
New generative artificial intelligence (AI) models such as ChatGPT are both revolutionary and controversial pieces of technology. They differ from other AI methods as they are able to produce content, including writing, images, and even songs. While generative AI has been in development for a long time, OpenAI’s ChatGPT stands out in its accessibility and capabilities. The model is free for anyone with an email account, and the base model version GPT 4 is available to companies that want their own chat bots. These models are very powerful, capable of producing content and art similar or even superior to that of a human. With these capabilities and ChatGPT’s sudden introduction, however, comes a lot of concern over how labor markets might react. Predictions and fear over AI automating traditionally human jobs have been present for a while, and ChatGPT’s capabilities have only increased them. In my technical thesis, my project explores an example where the introduction of AI could lead to the automation of previously human jobs. In my STS thesis, I focused more broadly on the implications of generative AI and labor based on recent events and trends.
The technical portion of my project focuses on work from a past internship. During the internship, I was tasked with finding a way to process and visualize the output of a satellite tracking explainable AI model, which included not only the satellite positions, but also various metrics used to determine risk. I created a Java GUI to display the output and give users more ways of interacting with the model. This was done through the Worldwind API and JFrame, which allowed an interactable globe and satellites. The metrics and explanations generated by the model were also displayed through JFrame and by using a simple language model. While the satellite data could be communicated clearly, the explanations generated by the language model were overly basic and limited information it could carry. Regardless, if this project were to be successfully implemented, it would automate satellite tracking and replace the human workers previously responsible. Future work that incorporates generative AI models such as GPT4 could advance the interactivity and trustworthiness of the model, making it even more viable to replace the human experts that have traditionally monitored satellites.
The STS portion examines the impact of generative AI in a wide range of industries including writing, tech, sales, and more. It compares previous thought notions of how AI will affect job markets with trends after the introduction of ChatGPT. The thesis focuses on the patterns exhibited by freelance markets, as it is a widely popular way that Americans find additional work and is representative of the labor market in whole. It was found that major freelance market Upwork has experienced an overall drop in demand and spending after ChatGPT’s introduction. It was also found that AI related jobs on the other hand have seen a very large jump in demand due to ChatGPT. This pattern is also seen in tech jobs, as many companies are laying off part of their workforce to expand their AI sectors. From this, it seems that much of the literature predicting how AI will interact with job markets have been correct, as there were some jobs severely affected while others have been created to offset the job loss. However, the new jobs are often more specialized, making it difficult for certain groups to attain. This isn’t a new phenomenon, but what is interesting about the new potential wave of AI related layoffs is that newly replaced are more white-collar workers, as compared to blue-collar works who were the most previously affected by technology.
Overall, my technical project presents an instance where generative AI can automate a job (satellite monitoring) while my STS project expands on general trends to gain a larger picture of the potential effect. The concrete example from my technical of an area automatable shows how wide-reaching generative AI can be, as satellite monitoring jobs aren’t very common and many wouldn’t believe them to be at risk. However, the fact that they could potentially be replaced raises concerns over how large the implications of ChatGPT actually are and whether those groups affected will receive enough attention. Similarly, many literatures cited in my STS project argued that those replaced by AI need protections. While the number of jobs lost will supposedly be offset by the new jobs created, the higher skill bar needed for the new jobs can equate to a large, negative social impact. This raises questions over what can be done to prevent this effect, and whether government regulations can help. Since ChatGPT has made an impact so quickly and so recently, this is an area that certainly needs to be studied more.
