A recent piece in Smithsonian Magazine, published just last week, looks back at the earliest attempts to teach machines, remembering a time when complex algorithms struggled with basic pattern recognition — like mistaking George Harrison for a woman. It’s a stark contrast to today’s sophisticated systems, which process vast data to optimize nearly everything. This drive for optimization, however, creates a real puzzle when considering human democracy. Writing just yesterday in Noema, HennyGe Wichers argues that democracy doesn't seek optimization or consensus; it needs friction, open contestation, and even conflict to truly function. Putting these perspectives side-by-side illuminates a core tension: how the historical trajectory of AI's learning, from simple pattern recognition to complex optimization, raises profound questions about its compatibility with the messy, conflict-driven nature of human governance. The story of the Perceptron, as told in the first piece, shows machines learning to refine their understanding through simple feedback, aiming for a singular, correct answer. When early AI misidentified George Harrison, the system was trained to correct that specific error, evolving to optimize its recognition. This historical arc of artificial intelligence, from basic pattern matching to complex optimization, is what makes HennyGe Wichers's argument so striking. Her piece on democracy doesn't just observe AI's application in politics; it questions the very premise of its "listening" capabilities. Where the Perceptron worked to eliminate "wrong" answers, Wichers points out that democratic systems thrive on disagreement. The Team Mirai chatbot, for all its success in surfacing voter concerns, still operates within "conversational boundaries" – a subtle echo of the early AI's structured feedback. The interesting thing is how the drive to optimize, inherent in AI's learning history, can clash with democracy's need for open contestation. The piece on the Perceptron shows a system designed to reach a clearer, more accurate understanding. Wichers suggests that applying this drive for clarity to democracy might inadvertently smooth over the very friction essential for its function. The evolution of artificial intelligence, from mimicking basic neurons to processing vast datasets, reveals a clear progression towards optimal solutions. Yet, human democracy often relies on imperfect processes, on the very friction and contestation AI is designed to smooth over. What’s left to consider is whether a system that inherently seeks consensus can ever truly coexist with the necessary messiness of collective human decision-making? Sources: Smithsonian Magazine: In the Early Days of Machine Learning, Massive Computers Said George Harrison Was a Woman. A.I. Has Come a Long Way Noema: Democracy Needs Friction To Function
