• D.C.
  • BXL
  • Lagos
  • Riyadh
  • Beijing
  • SG
  • D.C.
  • BXL
  • Lagos
Semafor Logo
  • Riyadh
  • Beijing
  • SG


AI polling company defends wrong predictions on the US election

Nov 6, 2024, 3:26pm EST
techpoliticsNorth America
AI illustration
Al Lucca/Semafor
PostEmailWhatsapp
Title icon

The News

An AI polling startup that tested its 2024 US presidential election model for Semafor, got most of its predictions wrong, but said that using artificial intelligence was still better than traditional polling.

Aaru — founded by two 19-year-old college dropouts and a 15-year-old — predicted that Kamala Harris would win the Electoral College. The company creates AI characters and feeds them a news diet based on their demographic and then asks them how they’ll vote.

But Aaru’s prediction showed Harris just barely edging out Donald Trump. “A coin flip is a coin flip; 53-47 is not significantly different from 48-52,” co-founder Cameron Fink told Semafor’s Reed Albergotti on Wednesday, after Trump won the presidential election.

AD

“Statistically speaking, we’re within the margin of error — so we did well,” Fink said. “That said, [there is] always room for improvement!” He also noted that Aaru’s prediction of Nebraska was almost perfect.

Title icon

Know More

“Like surveys of real people, Aaru got most of its predictions wrong,” Albergotti wrote. Major pollsters like Nate Silver’s Silver Bulletin and Split Ticket had put Harris as having better odds of winning the Electoral College, but the margins were razor thin.

Fink defended using artificial intelligence, saying AI polling is “significantly faster and cheaper than traditional polling, and still more accurate.”

AD

“If you’re going to pay for polling data that gets the wrong result, you might as well use AI and save money,” Albergotti wrote. ’“While surveying real people seems to be getting less accurate over time, the question is whether AI polling will improve.”

AD