Did a Human Write This?

On this episode of Galaxy Brain, Charlie Warzel talks to Max Spero, a co-founder of Pangram, an AI-detection company. They discuss how AI-detection tools work and how effective they can be at identifying what’s made by humans and what comes from a chatbot. They explore the cultural concerns around authenticity in the large-language-model era, and whether detection can keep up as models improve. The pair discuss how the speed of AI development and synthetic content threatens to degrade the quality of human writing and pollute the internet—and what, if anything, can be done to stop it.The following is a transcript of the episode: Max Spero: I want to see people using AI to cure cancer and, you know, make senior care easier and make all of our lives better. And I also don’t want to see AI polluting the internet. So sort of like: There’s these two sides, and I want to see the good side of AI flourish, and I want to help mitigate the harmful effects of AI as much as possible. [Music]Charlie Warzel: I’m Charlie Warzel, and this is Galaxy Brain, a show where today we are going to talk about the flood of AI writing online, and a tool that claims to help us determine what’s human and what’s been written by a chatbot.One of the big fears of the generative-AI moment is that we are entering this kind of slopocalypse. Chatbots, audio, video, and photo-creation tools—they all make it extremely easy for anyone to churn out synthetic content quickly. Especially text.By now you know the story: Students are using ChatGPT and other tools to write their essays for them. A rash of hastily self-published books on Amazon are clearly not real. Search-engine-optimization marketers and content farmers are flooding the internet with articles written by chatbots in order to game Google and make a quick buck. The web is just shuddering under the weight of all this