Impromptu: Amplifying Our Humanity Through AI * * * For the record, in case I fooled anyone, I asked GPT-4 to gener- ate this “article.” It’s fake news. Obviously, it’s nothing a human couldn’t write themselves. But it also took me mere seconds to produce—and it barely even scratches the surface of what’s possible. Indeed, of all the potential ways LLMs could negatively impact journalism, disinformation at scale tops the list. Right below that is the potential elimination of journalism jobs. Happily— if that word can be applied here—I believe the first possibility greatly reduces the chances of the second. How does this play out? Consider what Steve Bannon famously told journalist Michael Lewis in 2018: “The Democrats don’t matter. The real opposi- tion is the media. And the way to deal with them is to flood the zone with shit.” To date, the tendency of LLMs like Bing Chat and ChatGPT to produce inadvertent misinformation in their own faulty attempts to write statistically plausible “truth” has overshad- owed their potential use as disinformation fire hoses that bad actors could use to, ahem, muddy the waters of public discourse. Obviously, that potential is there, waiting patiently. When bad actors inevitably put these tools to such ends, I’m sure we’ll see calls for new regulation tailored specifically to AI-generated disinformation. We may even see efforts to ban 90