Machine Translation: A Comparative Analysis of Current Technologies; Globalized Tech Manufacturing and Workers Around the World: A Case Study of Nvidia GPUs

The basis of both my technical and STS research is examining the role of technology in an increasingly globalized world. The tie between my technical and STS research can be seen in these interconnected trends: increasing multinational tech manufacturing, increasing use of machine translation, and increasing demand for GPUs. My technical paper consists of a framework for comparative analysis of machine translation technologies. This was contextualized by the increasing importance of machine translation in daily multinational interactions, as well as by rapidly evolving techniques in the field of machine learning and their application to translation. My STS paper analyzes how workers in Nvidia’s GPU supply chain are being negatively impacted as demand skyrockets. The demand for GPUs is connected with increasing use of machine learning for tasks like translation, while translation itself is becoming more important with growing international supply chains like those for GPUs. Thus, my portfolio as a whole uses the technical and STS papers as lenses to examine two distinct but related facets of a currently evolving global sociotechnical system.

My technical report consists of a comparative analysis of current machine translation systems. Machine translation is a rapidly progressing field with many currently developing technologies, and it is important to determine which of these are effective and why. I aim to compare the strengths of these technologies to determine which are the most promising for future advancements. In the paper, I present a way for major current models such as Google Translate and ChatGPT to be evaluated on metrics of accuracy, flexibility, scope, efficiency, bias, and consistency. I then present reasons why I expect that all current translation technologies will have varying strengths and weaknesses. I conclude that depending on the intended application, the translation approach that ought to be used differs. Further conclusions could be reached by looking at a wider range of technologies and performing more rigorous and in-depth testing.

My STS paper analyzes how workers in Nvidia’s GPU supply chain are being negatively impacted and the core causes of these impacts. My literature review synthesizes information on several relevant factors: labor arbitrage, global supply chain structures, working conditions in tech supply chains, and rising demand for semiconductor chips. My analysis uses this information to argue about how rising demand for chips has led to the development of unstable employment structures in certain countries involved in tech supply chains, leading to negative outcomes for workers. These countries include China, Vietnam, Malaysia, Philippines, and several others in Southeast Asia. My conclusion gives some recommendations about how these effects might be mitigated or prevented in a long-term sustainable manner.

Writing both of these papers together was beneficial for my understanding of both topics and let me comprehend the associated issues in a more nuanced way than would have been possible doing each project in isolation. Each paper provided context to the other and let me see a more complete picture of the involved sociotechnical systems. For example, a key aspect of my STS paper was the recent rise in demand for semiconductor chips, which I argue played a role in harming working conditions in supply chains. However, my technical paper provides some context as to why that demand exists in the first place. Technologies like machine translation, which are increasing in importance, make heavy use of machine learning to function. In turn, training machine learning models requires extensive use of GPUs, which are fundamentally built around semiconductor chips, leading to increased demand. Furthermore, those same GPUs driving the demand are also responsible for degrading the working conditions surrounding their own construction. Thus, the technical paper provides important societal context that helps better understand the background to my STS research. In a similar way, the STS paper helps give insight into the broader societal implications of the technology driving my technical research. The GPUs underpinning machine translation technology are having a positive global impact through the technology they facilitate, but, as my STS research demonstrates, they also have negative impacts due to the circumstances surrounding their increasing production. In summary, working on these projects together has given me deeper insight into a complex sociotechnical system and has demonstrated how the technical side of an issue is inherently related to the social, and vice-versa.

School of Engineering and Applied Science

Bachelor of Science in Computer Science

Technical Advisor: Rosanne Vrugtman

STS Advisor: MC Forelle