United States
United Kingdom
Canada
Australia
New Zealand
IrelandNate Silver welcomes AI coders, but not AI voters
On Bloomberg's Odd Lots, the election modeler explains why synthetic respondents cannot replace the real people behind a poll.
Bloomberg Podcasts
Story of Why this story? The video opened at a pace of 37,754 views a day, 17 times a normal first day on this channel (2,240).
An Election Day Special With Nate Silver | Odd Lots
Things we can learn from the video
Press to jump
- 7:35Silver challenges assumptions about pro-Democratic polling errors and explains how errors correlate across similar electorates.
- 13:03House and Senate forecasts combine model projections, early voting evidence and the particular difficulties of contested states.
- 21:31AI-generated respondents lack the real-world evidence polling needs and can reproduce stereotypes instead.
- 25:17Aggressive gerrymandering can backfire, leaving its architects with fewer seats than the original maps would have produced.
- 37:00Wealthy traders can influence prediction-market prices, complicating comparisons with statistical election forecasts.
- 41:07Silver describes using generative AI to audit code and build statistical simulations, rather than replace survey respondents.
What does a poll lose when nobody actually answers it? For Nate Silver, the answer is its connection to reality. AI-generated respondents might reproduce stereotypes, he tells Joe Weisenthal and Tracy Alloway, but they cannot supply the real-world evidence that election forecasting needs.
That distinction runs through "An Election Day Special With Nate Silver | Odd Lots," which drew about 34,000 views in its first 19 hours, already far beyond Bloomberg Podcasts' usual first-day audience for regular videos.
Silver is not rejecting the tools. Later in the conversation, the Silver Bulletin founder describes using generative AI to audit code and build statistical simulations. His dividing line is between helping construct a model and inventing the people whose answers are supposed to inform it.
The election discussion also challenges the idea that polling errors always favor Democrats. Silver explains that errors can be linked across places with similar demographics, while midterms depend especially heavily on who turns out. A forecast is not simply a stack of survey results: pollsters make consequential judgments about which respondents are likely to vote.
Recorded as an Election Day preview, the episode considers Democrats' favorable political conditions alongside the difficulty of winning states such as Texas. Bloomberg's subsequent coverage reports Decision Desk HQ's projection that Democrats would take the House, a result update separate from this forecasting conversation.
Silver also turns to prediction markets, explaining how wealthy traders can influence betting prices. The useful question is not just what odds a market displays, but who has enough money to move them.
Published on Trari
Where next?
Is this your channel?For the channel's owners and team: a ready picture and text for your audience
Share the story with your audience. Everything is ready:
For an Instagram story or a WhatsApp status
For a community post or your channel's description
- Channel
- Bloomberg Podcasts
- Subscribers
- 609,000 ▲ 7,000 this week
- Channel views, last 7 days
- 2,619,446
- Channel views, pace per day
- 415,779 (on a normal day: 390,074)
- Video published
- November 3, 2026
- Video views so far
- 34,080 (less than a day online)
- Video views, pace per day
- 37,754
The numbers
The Odd Lots special had already exceeded the channel's usual full-day views well before its first day was complete.
💡 Test your knowledge
Who co-hosts Odd Lots with Joe Weisenthal?
Tracy Alloway and Joe Weisenthal co-host Bloomberg's Odd Lots podcast.