Daniele Carminati
Soft power is a term that is notoriously difficult to grasp resulting in being often misinterpreted and misused. The concept is often equated to cultural power, as the appeal of a country to foreign audiences in the form of lifestyle, cuisine, fashion, music, movies, and more. Although reductive, this interpretation gets to the general public. More precisely, Joseph Nye, the scholar who coined the term, describes it as “the ability to obtain preferred outcomes by attraction rather than coercion or payment.” This is hardly straightforward, and it may also be an overly positive way to understand dynamics surrounding the power of attraction. Soft power is not inherently good or bad, it is still a form of power that can be misused and abused. How does AI – as large language models (LLMs) such as ChatGPT – understand soft power? And why this question is more meaningful than it first appears.
These models rely on the information that is ‘out there’ (i.e., the Internet), hence they generally combine and summarize already available knowledge, but they are currently less effective in creating new ideas. However, when pushed beyond parroting definitions found in the literature, they can reveal insightful patterns resulting from the discourse surrounding the term, as a form of ‘collective knowledge’ and understanding of this elusive concept. In other words, LLMs can quickly connect the dots, and the results, as this article aims to show, can be as insightful as concerning.
No comments:
Post a Comment