Exploring AI and U.S. Politics: Insights from Processing 80 Million Tokens with ChatGPT-4

Greetings, readers! Four months ago, I embarked on an intriguing investigation aimed at understanding the intersection of Artificial Intelligence and U.S. politics. My objectives were to explore a few key inquiries:

1. What perspectives does AI hold regarding U.S. political dynamics?
2. Is AI capable of summarizing and analyzing political legislation? If so, what might those interpretations reveal?
3. Can these insights be valuable for legislators and policy makers?

In pursuit of answers, I utilized ChatGPT-4 to analyze the full text of an astounding 13,889 congressional bills from the 118th session, ultimately processing around 80 million tokens. The results were both enlightening and unexpected.

Key Findings from the Analysis:

1. Liberal Perspectives Emerge: It became apparent that ChatGPT-4 often exhibits strong liberal tendencies, frequently advocating for social equity and the empowerment of marginalized communities.

2. Optimizing Data Processing: A significant takeaway from my project was the effectiveness of OpenAI’s Batch Processing API for managing extensive data sets. By leveraging this tool, I managed to process nearly 40 million tokens in just 40 minutes at half the cost of alternative methods.

3. Competence in Bill Interpretation: The AI demonstrated impressive capability in interpreting complex political bills. For instance, when analyzing this specific legislation, it not only grasped the purpose of mifepristone and its medical implications but also recognized misleading claims, emphasizing the detrimental societal effects of spreading misinformation. Such insights from an algorithm are remarkably profound!

4. Agreeing with AI’s Analyses: Interestingly, I found myself in agreement with many of the AI’s interpretations. For example, the analysis of another bill (this one) initially struck me as questionable. However, upon reflection, I began to appreciate the nuance of the situation, suggesting that there’s a legitimate concern regarding potential adverse reactions to the COVID-19 vaccine, raising the question of whether government intervention might indeed be warranted.

5. Detecting Legislative Corruption: Regardless of one’s stance on specific legislation, I am confident in asserting that AI excels at identifying overt corruption among legislators. The data suggests that the AI’s aggregate interpretations—based on over 1,200 bill analyses