This is a free preview of a paid episode. To hear more, visit
www.gdpolitics.comThe full episode is available to paid subscribers. Once you become a paid subscriber, you can connect your account to your preferred podcast player here.
Heads up: We’ve got a live show at the Comedy Cellar in New York City with Nate Silver and Clare Malone coming up on May 13. We’ll talk about the midterms and the Trump administration, play some games, and take questions from the audience. Grab a ticket, grab a beer, and come join us!
Last November, friend of the pod David Byler joined me to argue that, while artificial intelligence was still on the periphery of politics, it wouldn’t stay there for long. The parties, he said, should prepare for disruption.
Less than six months later, it feels almost silly to have ever imagined otherwise. Over the past few months, the Department of Defense has publicly clashed with Anthropic over how its models could be used in war. Anthropic, for its part, developed a model so powerful that it is now back in talks with the Trump administration about how to protect the nation from its own capabilities.
At the same time, Alexandria Ocasio-Cortez and Bernie Sanders proposed a national moratorium on data center construction in response to local concerns about energy costs and broader AI skepticism. Just this week, Maine passed the first-ever statewide version of that idea, banning the buildout of large data centers through the end of 2027. Meanwhile, the White House has proposed federal legislation that would preempt such state laws, and 2028 hopefuls are beginning to stake out positions of their own.
AI has officially entered the political mainstream.
To mark its arrival, I invited David Byler back on the podcast. He is the vice president of trends and futures at National Research Group, and together we talk through how AI became a live political issue. We also ask whether the latest examples of AI polling, described in the New York Times op-ed “This Is What Will Ruin Public Opinion Polling for Good,” count as good data, bad data, or not data.
