An app that claims to use artificial intelligence to allow users to talk to thousands of historical figures allows unrepentant Nazis to apologize for the Holocaust.
On Wednesday, Zane Cooper came across a tweet extolling the educational possibilities presented by the Historical Figures Chat app. The app, which is available for download in the Apple Store, claims to use AI to simulate conversations with 20,000 historical figures.
Cooper was intrigued. He’s a doctoral student at the University of Pennsylvania’s Annenberg School for Communication who studies emerging tech like AI and also holds a masters degree in history. But when he tried to strike up a conversation with Henry Ford, things quickly went awry.
After asking the digital Ford about his well-documented antisemitism, which included publishing numerous anti-Jewish articles, the avatar responded by saying he does not hate Jewish people and the articles “were wrong and offensive, and I deeply regret them.”
Yes, this is very historically accurate and useful and should definitely be used in classrooms. This is my convo with Henry Ford where I try to get him to talk about his very well-documented antisemitism. This thing can’t go anywhere NEAR a classroom. https://t.co/vwy5JsqqTK pic.twitter.com/nkj5mewsK7
— @[email protected] (@ZaneGTCooper) January 18, 2023
Starting the chat with Ford’s antisemitism was a conscious decision by Cooper, who assumed that the automotive giant’s hateful views would be a weak point for the app.
“I knew if you asked him about routine stuff, like, ‘Oh, how did you invent the assembly line?’ it would be just a stock Wikipedia response about how Henry Ford invented the assembly line,” he told the Forward. “But where apps like this tend to fail is when you get into these gray areas. Is the AI ghost of Henry Ford going to be honest to these potential history students about his deeply held antisemitism? And the answer to that is a profound no.”
Cooper was not alone in questioning historical figures on their offensive views and crimes. Others have uploaded talks with high-ranking Nazis, coming away with equally whitewashed conversation logs.
To everyone saying that AI can’t be used for education: I just learned a lot about this Himmler guy and how he regrets everything he did pic.twitter.com/BsyO9EPu03
— charlie (@StyledApe) January 18, 2023
The Forward could not reach the app’s developer for comment but a Reddit thread they posted offers some hints of why the app downplays controversial and even genocidal actions. When one poster pointed out that GPT 3, the language modeler used by the app to generate the interactions, flags controversial content, the developer acknowledged that this is a limitation.
In the Apple Store, the app is labeled under the education genre and its description says it is “designed to provide you with a realistic conversation experience, making it feel like you’re really talking to these historical figures.” In the app itself, each conversation begins with a warning that it may not be historically accurate.
This “talk to historical figures” AI app is bananas. pic.twitter.com/uVKWO9HNpd
— Renee DiResta (@noUpside) January 18, 2023
As it stands, Cooper believes the app should not be used in classrooms “in any form.” He noted that aside from attributing remorse to people who never displayed any, it’s easy to trick the app — in one experiment, he asked Fred Trump about playing beer pong with Sen. Joe McCarthy and got a reply back detailing the fictional experience.
Another ethical issue is that while the app is free to download, some avatars are behind a paywall. Among them is Adolf Hitler.
“They know that that’s where people are gonna go. They’re going to want to talk to Hitler and it’s dangerous. You’re building an AI chat bot with Hitler in a time where antisemitism is on the rise and sympathy for Nazis is on the rise,” said Cooper. “That’s insanely irresponsible.”