Deepfakes have productive enterprise uses too

Image created by DALL-E

Deepfakes aren’t always bad.

There are several good uses of the technology in the enterprise sector. To go back to the previous discussion and sum it up, deepfakes are digitally manipulated video or audio recordings that replace someone’s face or voice with that of someone else, in a way that appears real – and this makes businesses vulnerable to scams. It is made possible through technology called deep learning.

While deep learning has been used to create synthetic media for dubious activities like disinformation, fraud, and corporate espionage, the technology’s positive use cases need to be considered to evaluate it. According to John Shier, Field CTO Commercial at Sophos, the technology can be used for upskilling employees in a more cost-effective way.

“Let’s say you’re Subway, the sandwich shop. It has hundreds of thousands of employees speaking different languages and you need to produce training for them. You can hire actors that speak local languages or translate it into different languages. With deepfake technology, you can create the training, specify what it’s going to say, identify which actor to hire, then generate it in 100 different languages. You can do that at scale and at much less cost,” Shier explained.

The infrastructure for breaking down language barriers is already in place. English football icon David Beckham, for instance, spoke in nine different languages through AI to promote his Malaria No More campaign. This was carried out via a process called visual synthesis.

Synthesia, the company behind Beckham’s project, has also devised a method called “Native Dubbing”, which can be used for translating and localising video content, but on a larger scale.

Training AI

Meanwhile, at identity verification specialist Onfido, deepfake technology is used to stress test biometric systems and train their AI to recognise anomalies in video and video feeds, said Albert Roux, the company’s VP of Product Management for Fraud.

“It has been used to better assist those with accessibility problems,” Roux added. “Microsoft’s Seeing AI and Google’s Lookout leverage AI for recognition and synthetic voice to narrate objects, people, and the world. AI-generated synthetic media can power personalised assistive navigation apps for pedestrian travel.”

Rick McElroy, Principal Cybersecurity Strategist, VMware. Image courtesy of VMware.

Rick McElroy, Principal Cybersecurity Strategist at VMware, noted that the technology can be useful in creating synthetic data sets for training machine learning models.

“Deepfake technology can foster accessibility and create learning tools for the healthcare and education sectors – for example, recreating people and scenarios in history or rendering fake patients to be used in research. This protects patient information and autonomy, while still providing researchers with relevant data,” said McElroy.

This, however, is not a pie-in-the-sky concept; synthesising data is already being explored by Nvidia, MGH & BWH Center for Clinical Data Science, and the Mayo Clinic, where researchers created fake brain MRI scans to train algorithms in spotting tumours more effectively.

The technology can also turn customers into models to test new products or features, said Gary Davis, Chief Cybersecurity Advocate at BlackBerry. This, said Davis, protects customer autonomy while providing researchers with relevant data.

Albert Roux, VP of Product Management for Fraud, Onfido. Image courtesy of Onfido.

Turning back time

Another industry benefiting from such synthetic media is entertainment. “The film industry has used deepfakes to great effect, helping bring actors back to life to complete movie series and to place an actor’s face on stunt doubles,” remarked Onfido’s Roux.

Some recent examples include the following:

  • AI firm Metaphysic made a deepfake performance of Elvis Presley on the stage of America’s Got Talent.
  • Actor William Zabka was de-aged in the TV series Cobra Kai so he would look like his younger self in the 1984 film The Karate Kid.
  • Deceased rapper Tupac Shakur returned to the screen in a Snoop Dogg music video through an intricate deepfake.

The technology, said VMware’s McElroy, has also been used in news outlets to generate the voices of popular podcasters or radio hosts in different languages.

Other applications

Other areas where deepfakes can have positive applications are customer support, as well as in cultural institutions.

John Shier, Field CTO Commercial, Sophos. Image courtesy of Sophos.

“Instead of having call centres where people speak all sorts of different languages, you have a machine that responds in the caller’s language, and can have a conversation with that person. This can replace the frontline so that as you’re triaging customer support calls, you’re then able to get the right experts on the line afterwards,” said Shier, who estimates that such a setup can probably handle about 80% of the basic support calls.

Museums, on the other hand, can use deepfake technology and hold interactive exhibits where people can watch a video or immersive AR or VR experiences. The visitor, for example, can stand virtually in the Colosseum in Rome, watch gladiators fight, and see Caesar give the thumbs up or the thumbs down.

“These exhibits would be really interesting from a cultural and just cultural point of view,” remarked Shier.

The Salvador Dalí Museum in the United States, in particular, has already created something similar: an interactive, life-size deepfake of the eponymous artist, which takes selfies with visitors.

Gary Davis, Chief Cybersecurity Advocate, BlackBerry. Image courtesy of BlackBerry.

Marketers can also benefit from deepfake, said BlackBerry’s Davis. “Instead of an in-person actor, companies can purchase a licence for an actor’s identity and craft an omnichannel campaign with appropriate speech. In this manner, deepfake is gaining traction as a way for companies to engage customers and deliver value.”

Like any piece of technology, deepfake is a double-edged tool. Bad actors can use it to defraud – and they have. But organisations have also leveraged it to reduce costs, instruct employees, advance healthcare, and improve processes.

To stay on the path of using deepfakes constructively, enterprises should consider its ethical implications and wield the technology responsibly. As deep learning continues to improve, businesses must stay informed about the capabilities and limitations of deepfakes, and to use it in ways that benefit both the company and society as a whole.