Voice-cloning technology bringing a key Supreme Court moment to ‘life’

NEW YORK — Seventy years ago Friday, no one outside the U.S. Supreme Court building heard when Chief Justice Earl Warren announced the landmark Brown v. Board of Education decision on school desegregation.

Using an innovative voice cloning technology, it will now be possible for people to “hear” Warren read the decision as he did on May 17, 1954, along with oral arguments from attorneys, including future Supreme Court Justice Thurgood. Marshall.

The “Brown Revisited” recreation should be available Friday at brown.oyez.org, a website that was the dream of former Northwestern University professor Jerry Goldman. Goldman painstakingly crafted the site so people can hear oral arguments in decades of Supreme Court cases and follow along with written transcripts. Yet Goldman has always been frustrated that the court did not begin taking oral arguments until 1955 – a year after the Brown decision was handed down. Printed transcripts just aren’t the same.

“I could give you the libretto of ‘Madame Butterfly,'” he said. “But would you rather read it, or sit and listen to the performance?”

The Brown decision was a milestone in the civil rights movement. The court overturned an 1896 decision that institutionalized racial segregation with “separate but equal” schools for black and white students, ruling that such accommodations were anything but equal.

While the court began recording arguments in 1955, virtually no one heard them until 1969, when they were made available for scholarly and legal research through the National Archives. Full public access was not granted until 1993. The court began posting arguments on its website in the 2000s, but usually with a delay of several days.

It was only in 2020 that the court made regular livestreams of the oral arguments available. Cameras are never allowed.

A year ago, Goldman said, he attended a play in which artificial intelligence was used to imitate a famous voice, and he wondered if this technology could be used for historical court arguments. One Northwestern alum, James Boggs, CEO of the interactive audio company Spooler, expressed interest when contacted.

“It’s good to draw attention to this case,” Goldman said, “because it is fundamental to our understanding of the Constitution and has changed America.”

The first step was to find recordings of the long-deceased clients in the business, preferably made around 1954 to approximate what they sounded like at the time. That wasn’t difficult in the cases of Warren, a former governor of California, and Marshall. It was more difficult for integration opponent John W. Davis, whose long career included becoming the 1924 Democratic presidential candidate. He died in 1955.

A Davis recording was tracked down through the Library of Congress. Recordings of some other participants could not be located.

Using artificial intelligence, these voice clips were fused with those of actors reading the historical transcripts to make it sound as if they were speaking again.

The actual fights were spread out: 18 hours over three days, with 38 participants. Goldman boiled things down to an hour and 45 minute presentation, including Warren’s reading of the decision. Goldman consulted the written notes Warren had left behind, allowing the game to include the chief justice’s emphasis that the decision had been made unanimously.

Technology’s growing ability to recreate votes is a miracle, yet deeply disturbing to many who fear it could put false words into familiar mouths – such deepfakes are a particular concern in the run-up to the presidential election.

Ravit Dotan, CEO of TechBetter and an instructor on the ethics of technology, said she is concerned about the practice of cloning people’s voices without their consent, even though consent is not possible from people who are no longer alive. She believes “Brown Revisited” sets a bad precedent.

“In the future, I can imagine laws determining how long a person’s likeness rights survive after their death, similar to copyright, which expires 70 years after the creator’s death,” Dotan said. “But at the moment there is no legal guidance, and I worry that people are taking advantage of that, exploiting people’s likeness or even spreading misinformation.”

Rather than a deepfake, the Brown project is a “deep truth,” Boggs said.

“We are not creating new content,” he said. “These were things that were actually said and we have the historical documentation to prove it.”

Similar recreations have a natural limit. It was not until the end of the 19th century that sound recordings of voices were available. Go back further, and they would essentially be guesses. Who knows what George Washington actually sounded like?

But for the curious, the ‘Brown Revisited’ project offers a new window on history.

___

David Bauder writes about media for The Associated Press. Follow him at http://twitter.com/dbauder.