I keep a spreadsheet of great thinkers. From Homer to Hume. From Charlotte Brontë to Ta-Nehisi Coates.
This morning, I was recalling a thought-provoking photo of Magritte’s art I had stumbled upon when browsing Twitter last night. I went searching for it. Searching Twitter and my browser history like a London man having lost his hat to the night wind, somewhat hopelessly. I had not saved it in the moment, trying to live in the moment rather than catalog it. I let this inspiration lead me. I decided to reopen my spreadsheet, fire up Wikipedia, and learn more about Magritte.
As I began to do this, I realized I had not entered a field yet for Magritte. Born in 1898, I chronologically slotted him in between Marcus Garvey (1887) and Leo Strauss (1899), who distinguished scholars from thinkers, the latter solving the world’s great problems, with the former merely studying the differences between such thinkers. I began to type “René…” into a newly created row. “René Descartes” popped up as autocomplete given it was a prior entry. The technologists at Google succeeded in creating such a frictionless experience for the majority case, I hit “Enter”. Before I knew it, Magritte was now Descartes. Indeed, he was not, nor has the world quite seen two René Descartes that would warrant inclusion in this list. But technology thought it was helping me when in fact, there is a much more sinister and harmful trend occurring before our noses in the name of efficiency and ease of use.
Technology builds efficiency by learning. Consider Google’s writing suggestions in emails. Consider Facebook, Twitter, and Linkedin recommended connections based on who you already know. Consider GPT-3’s generation of text based on existing information collected and then artificial intelligence generating what it predicts to be the right text. None of these systems are designed for new input.
Creativity and innovation often come from breaking boundaries. Writing something in a document that doesn’t conform to the prior standards or sentiments. Meeting someone who has no mutual connections to you is at the heart of serendipity. These experiences expand our horizons, rather than constrict them. The filter bubbles of social media may just be one harbinger of thought bubbles and social bubbles that the tools we’re building could otherwise break.
The more that systems drive efficiency through repetition, the less we are likely to even see the road not taken.
Love your point that machine learning systems aren't designed for new input - things like GPT-3 just remix what already exists. Some people think humans operate the same way though - that no create act is truly novel. Who knows. But regardless, agreed that algorithms are probably shrinking the creative window.
That aside, these recommendations and autocompletions are especially pernicious because they seem so small and benign, to the point that we barely notice them happening. What's the big deal - you clicked on Descartes instead of Magritte today. But as we get hundreds (thousands?) of these algorithmic recommendations and autocompletions a day, the impact is very material.