Toronto Star February 25, 2024
LOS ANGELES—One recent day in Los Angeles, Holocaust survivor Pinchas Gutter answered questions about his family’s murder by the Nazis.
Sitting comfortably among a small audience, Gutter’s eyes grew sombre as he described his parents’ murder, his back-breaking work as an 11-year-old slave, and eating insects to survive. He recalled with emotion his last glimpse of his twin sister’s blond braid before she vanished forever.
But when asked how he felt about the present war in the Middle East, Gutter’s expression went blank and he was unable to answer.
That’s because Gutter was not in Los Angeles in 2024 but was speaking as a life-sized artificial intelligence image recorded in 2014 at the University of Southern California Shoah Foundation Museum.
The real Gutter, a 91-year-old Holocaust educator and frequent guest lecturer, lives in Toronto and was the first survivor recorded in interactive AI by the USC foundation, which was founded by Steven Spielberg, for its innovative Dimensions in Testimony.
The project, which is racing against time to preserve the memories of survivors who are now more than 80 years old, uses advanced recording techniques and natural language processing to capture survivors’ stories in extensive interviews. AI technology then enables users to interact with a life-sized image of Gutter that can be questioned and will respond much as if talking to the real person.
“I tell my story for the purpose of improving humanity drop by drop by drop. Like a drop of water falls on a stone and erodes it,” Gutter’s AI image said, “to make the world a better place.”
For educators and technology experts who spoke with the Star, Holocaust memorial projects, like USC’s, are an important example of AI’s transformative force for good.
But the advantages of AI and evolving technology for historians may also be its Achilles heel. As much technology can help to bring the past to life, it can also distort and twist it to devastating effect.
From its beginnings as a mid-20th-century theory that machines could use human-like reason to solve problems and make decisions, AI now enhances many aspects of our daily lives, from virtual assistants like Siri and Alexa to recommendation systems on streaming platforms like Netflix to advanced medical diagnostic systems.
But AI deepfake videos — which distort images to feature real people saying or doing things they never did — are now easy for cybercriminals to create and use to sow chaos on facts.
“Before you needed a PhD and fancy graphics and video editing software” to make deepfake videos, said Jimmy Lin, a professor at the David R. Cheriton School of Computer Science at the University of Waterloo. “But now you just download an app.”
In January 2023, AI startup ElevenLabs launched a tool giving users the ability to create new synthetic voices for text-to-speech audio or clone somebody’s voice. Just days after the tool went live, a video of former child actor Emma Watson reading Adolf Hitler’s “Mein Kampf” and wearing a Nazi uniform was circulated on 4chan, an anonymous message board known for spreading hateful content.
Similar deepfakes have already marked the U.S. presidential election, with robocalls of an apparent AI-generated deepfake of President Joe Biden’s voice discouraging voters from attending the New Hampshire Democratic primary. Deepfakes are also expanding the reach of pornography with multiple free websites capable of “undressing” women in just minutes, according to Daniel Siegel, a graduate student at Columbia University who researches how AI is being exploited for malicious purposes.
Recently there has also been a proliferation in the use of AI image generators to negatively depict Jewish people in relation to the war in Gaza, said Siegel.
Muslim communities have also been rocked by anti-Islamic sentiment since the Oct. 7 Hamas attack, but it is antisemitism that has been the focus of hate-filled, far-right groups using AI tools, according to Memetica, a threat analysis company that monitors online hate.
“People who spread hate (online) reach audiences better than we do,” said USC Shoah Foundation executive director Robert Williams, in a panel discussion last September. “We need to figure it out, and we need to figure it out quick.”
But many organizations like the Shoah Foundation are already hard at work doing just that. With most Holocaust survivors now in their 90s, archivists, historians and museums are turning to immersive technologies to preserve their stories for future generations.
For instance, Toronto’s Holocaust Museum and the National Holocaust Monument in Ottawa are leveraging technology to customize first-person accounts, artifacts, photos and documents and to offer virtual reality experiences.
These immersive exhibits “help people see themselves reflected in the history,” said Michelle Fishman, director of marketing and communications for the Toronto Holocaust Museum.
Making survivor stories interactive preserves important historical truths, bridges the gap between generations and ensures that the lessons of the past remain accessible and engaging for future generations. And leaders of these projects as well as technology experts told the Star that while more protective measures are still needed, the benefits of using AI to preserve history still far exceed its dangers.
Survivors themselves seem to have little fear about embracing technology to preserve their memories. The USC Shoah Foundation has collected 56,000 audio-visual Holocaust testimonies in 44 languages from 65 countries.
Building on this success, the foundation has now expanded its reach to educate on the horrors of other genocides by collecting testimonies from survivors and witnesses of the Armenian genocide during the First World War, the 1937 Nanjing Massacre in China, the Cambodian genocide of the 1970s, the Guatemalan genocide of the early 1980s, the 1994 genocide against the Tutsis in Rwanda, the ongoing South Sudan civil war, the Central African Republic conflict of 2012 and anti-Rohingya mass violence of 2017.
In a related project, the final shattering of Jewish life in Germany on Kristallnacht, “the night of broken glass,” is being memorialized through virtual reality by the Holocaust Claims Conference in partnership with USC and Meta.
The immersive experience, to be released later this year, features Kristallnacht survivor Charlotte Knobloch, 91, who walks users through her ravaged Jewish neighbourhood in Munich as she did with her father on Nov. 9, 1938, when she was six. Viewers will experience first hand how ordinary citizens turned on their Jewish neighbours, burning synagogues, looting Jewish businesses and vandalizing their hospitals, homes and schools.
For Knobloch, former president of the Central Council of Jews in Germany, fighting antisemitism through projects like this one is a matter of life and death. Speaking in German in a Zoom interview with the Star from her Munich home, the great-grandmother said she has needed a security detail for many years because she has had multiple threats on her life by neo-Nazis.
“The security situation is worse than ever, and it is very scary,” she said.
Knobloch, who survived the Holocaust by hiding with a Catholic family in a small town, never forgot her Jewish roots and devoted herself to rebuilding a Jewish community in Germany.
“We need the commemoration to prepare for the future,” Knobloch said. “In five to 10 years, this generation will be gone. And it will be up to our descendants to … make sure that we don’t forget these horrible crimes.”
The advent of AI recording and storytelling by actual survivors is likely to make it much easier to reach new generations with the truth.
Although modern antisemitism and Holocaust denial are deeply painful for survivors, it often drives them to become involved in the projects that serve as an antidote to that denial.
To protect historical testimonies, scholars and technology companies are finding ways to authenticate genuine content and detect deepfakes.
Startup companies such as Reality Defender are scanning user-generated media to detect deepfakes and providing alerts to deepfake image manipulations, said Columbia University’s Daniel Siegel.
Industry alliances such as the Coalition for Content Provenance and Authenticity are also helping by creating standards to certify the source and provenance of media. This allow creators to create content that can’t secretly be tampered with, Siegel said.
Working with these standards, Meta, an industry-leading provider of social media, committed to a major boost in protecting historical truths, in a Feb. 6 announcement. The company said it will detect and label AI-generated images that users post to its Facebook, Instagram and Threads sites and could apply penalties to users who fail to label their own altered audio and video content.
“When it comes to harmful content, the most important thing is that we are able to catch it and take action,” the statement said.
Waterloo’s Jimmy Lin sees the importance of these controls and oversight along with education to help people navigate and critique online information. If people can learn to consider who is behind AI sources and reflect on different points of view, they will be less influenced by deepfakes, he said.
But there is no doubt for Lin that the ability to have an AI conversation with a Holocaust survivor like Pinchas Gutter is one of the best ways to counter deepfake attacks.
“This is a really novel and forward-looking use of generative AI technology. I think this is refreshing.”
Pinchas Gutter agrees. He hopes using technology to share his story will help build tolerance and understanding in the world, he said in an online chat this week.
“We must … accept different cultures and different religions and different ways of life, he said.
“I say Psalms, you read the Qur’an,” he said. “Why can’t we live together?”
At its best, AI’s ability to keep truth powerfully alive through the experiences of witnesses may make that more possible, not less.
View the full article on the Toronto Star website here.
