LESSON 17

When we were kids, Americans largely agreed on what the facts were. Political disagreements were generally about what to do given these facts. No longer.

We can trace this process back to the end of FCC’s Fairness Doctrine, the subsequent rise of cable news, the social media revolution, the filter bubbles that resulted — and the ways in which political rhetoric has adapted to this new landscape. Talking politics at the family reunion has become a recipe for disaster.

Bad as this all may seem, it could get a lot worse.

OpenAI and other tech companies are pushing to develop individualized, personalized large language models that serve as intermediaries between us and the information on the Internet.

Rather than interacting with web content directly, the idea is that we would each have our own personalized information concierge, an AI model that finds information specifically tailored to our interests, beliefs, and values, and then repackages this information specifically for our individual consumption.

Learning from the vast amounts of personal information that you generate every day as you navigate the internet, these personalized LLMs will deliver you fully-written AI-articles about your local community, spun to reflect your political leanings. They'll regale you with AI-generated podcasts about your latest hobbies, your favorite artists, the sports teams you follow. And they'll feed you conspiracy-theories designed to reinforce your personal suspicions, whatever they may be, about shadowy forces controlling our world.

Yes, you'll be getting individually tailored content. But you'll also be reading arguments that no author cared enough to take the time to make and listening to conversations that no speaker cared enough to actually have.

You'll also converse with them. Because users don't like combative or even stubborn chatbots, they'll agree with you when you're wrong, humor your misunderstandings, reconfirm your misperceptions, aid you in motivated reasoning, and generally fail to hold you accountable the way an actual human would. You'll get used to that. You'll become confident in your judgment and intolerant of being contradicted.

Now, maybe these personalized AIs won't take off. Maybe the human desire for authenticity (Lesson 14) is too strong. But if they do take off, our ability to make collective sense of the world will be further fractured. We won’t inhabit one of several competing filter bubbles, each of us will be in our own private filter bubble.

And of course, just like everything else on the Internet, these systems will be distorted and manipulated to sell us junk while jealously hoarding our attention. So we will each inhabit our own filter bubble tailored, not even to our personal preferences, but rather to the preferences that corporations have for us as individual marks for their grift.

At that point, every one of us will live in a private Truman show, a simulacrum created wholly for one person. As a society we will be utterly incapable of making fruitful collective decisions because we will have no collective understanding of the world.

Unknowingly inhabiting a Truman Show, your every move watched without your knowledge — that is definitely nightmare material. But even more existentially alienating is to live in a Truman Show where no one but an AI is watching.

Note: This lesson is based on a 2024 Scientific American essay by Carl Bergstrom and Brandon Ogbunu.