Deep fakes have been making headlines for all the wrong reasons. But can they be used for good purposes as well? The answer is Yes. Deep fake technology can be very powerful and can be used in many productive ways too. In fact it will increasingly have a constructive role to play in many fields like Ecommerce, healthcare, art, history, etc.
Like with any other technology, there are bad actors and good actors, bad uses and good uses. Deep fakes are mostly known for their vicious uses. You've probably seen them already. For example, realistic videos of politicians and celebrities saying outrageous things that they never intended to say.
Deep fakes have even been used as weapons of war and disinformation. During the first weeks of the Russian invasion of Ukraine, a deep fake of Ukrainian President Volodymyr Zelenskyy where he appears to ask his troops to surrender circulated on social media and was placed on a Ukrainian news website by hackers. Similarly, Jordan Peele made a deep fake video of former US President Barack Obama to precisely make a point about the dangers that deep fakes may pose in spreading political misinformation.
Most of the focus though has gone to these negative aspects of the technology. While all these concerns are more than valid we should also not lose sight of how we can leverage this technology for good. Many times you can find gold if you dig out the dirt. Here are some of the ways deep fake technology can be used in positive ways.
4 ways deep fakes are being used for good
There's a compelling case to be made for the proper commercial use of deep fakes: to allow personalization to happen at scale. Personalization is a good thing but to make it happen at scale is a problem that many businesses face today. Video Personalization is thus the business use case of deep fake technology. Businesses and their influencers want to be able to connect with their customers or fan bases in a more personal way, especially through mediums like video that are so powerful, and video personalization adds an extra layer of effectiveness. Without AI and deep generative models (the same tech behind deep fakes) this would be impossible to achieve at scale.
Some may question if this would be the authentic thing to do. When used maliciously, deep fakes over-impose something that the individual never intends to say or do in a video. This is definitely not authentic and it's ill-intentioned. But there's a clear difference between that and when you actually intend to say every single word to every single recipient of your video. Let's say a business or an ecommerce store founder wants to send a personal video message calling each customer by name. It becomes an impossible task when there are hundreds of customers or fans they'd like to greet or personally thank.
This is where the technology can come in handy and finally allow you to execute these great gestures. Customers want to feel special, they want to see the faces behind the brands, they want to connect with the real people behind the products that they are enthusiastic about. Ecommerce founders and influencers should equip themselves with the tools and technology that enable them to connect and interact personally with each one of their customers to ensure they all have a great experience and feel appreciated. How you are able to execute it will not matter more than the good intention behind the gesture.
Maverick is a great example of how you can achieve video personalization at scale through AI and deep fake technology. Maverick helps businesses especially in the ecommerce space to send personal video messages to all their customers at scale. You only record once and the AI can generate multiple unique versions of the base video where you call each customer by their name. This is a tool that enables ecommerce stores to finally connect with their customers in a more human way and allows them to act on these nice personal gestures at scale. Each video personalization represents exactly what you want to say to each customer.
2. Medicine and healthcare
Deep generative models can be also harnessed to have a positive impact in healthcare. Medical researchers often run into issues when trying to spot medical conditions using AI because of lack of data due to privacy reasons, financial constraints, and rare diseases with few examples and little data to pull from. Generative adversarial networks (GANs), also used to create deep fake videos, allow them to create realistic-enough medical images for their AI to learn and be trained on. This can lead to better ways of diagnosing and further treating diseases.
GANs have a lot of potential in medical research. According to the UCL Centre for Artificial Intelligence, hospitals can now collect and use their small samples of historical real patient data and use it to recreate an entirely imaginary population of virtual patients, without having to use and share the data of their live real patients.
In 2018 the Mayo Clinic collaborated with Nvidia and the MGH & BWH Center for Clinical Data Science to create 'fake' brain MRI scans using GANs. They found that they could get their algorithms to spot tumors by training them on these 'fake' medical images and on 10% real images removing the need to collect thousands of real images which can be expensive and daunting.
Deep fakes could also be used in other practical medical settings to help patients who have lost motor, speech or visual abilities (among other impediments and disabilities) to communicate better. This can help medical professionals and family members understand their needs better. They can give the power of better self-expression to people who are unable to do it themselves.
3. Art and History
Deep fake technology can help us recreate the past to be able to connect with it and understand it better. It can allow us to experience things that existed before our times and that we wouldn't be able to comprehend otherwise.
Recently a Scottish company, CereProc, was able to recreate audio of the speech that former US President John F. Kennedy was due to give in Dallas the day he was assassinated. They trained their own deep fake algorithms on audio recordings of the former deceased president to recreate the speech with his own voice.
A similar project from the MIT Center for Advanced Virtuality recreated a speech that thankfully was never delivered by former President Richard Nixon. The speech was prepared in case the moon landing mission may have failed in 1969. These projects help us learn what our leaders were thinking and how they were preparing for important events like this.
For a special exhibition some years ago, the Dalí Museum in St. Petersburg, Florida used deep fake technology to have a life-sized recreation of Salvador Dalí (famous surrealist painter) say a number of quotes which he actually wrote during his life. The AI was trained thanks to thousands of hours of the artist's old interviews.
Similarly, AI researchers at Samsung Labs have developed an AI that can create high-resolution avatars from a single still image or painting. They were able to have the Mona Lisa come to life and move in a very natural way allowing us to experience the painting in a completely new way.
4. Media and entertainment
Reuters came up recently with the first synthesized news reports using deep fake technology. They were able to recreate new video reports using data from previous video recordings from popular news presenters. The main innovation here is that they were able to generate news reports that were personalized for each individual viewer as if the popular reporter was presenting them exclusively to each viewer.
Moreover, a UK-based NGO called Malaria No More UK used deep fake technology to create a professional marketing video where well known soccer superstar David Beckham appears delivering a message in nine different languages to raise awareness about malaria and its consequences. Thise video personalization garnered a lot of media attention and was able to be reproduced in many countries around the world in the different languages it was produced.
Nvidia has been also leveraging deep fake technology to improve graphics for video games at scale. Just Eat was able to localize a video commercial featuring Snoop Dogg in Australia thanks to deep fake technology. The original music video which was highly successful in Europe mentions Just Eat but its subsidiary in Australia is called Menu Log. With deep fake technology the video commercial that was shown in Australia adapted Snoop Dogg's video and voice to say Menu Log instead of Just Eat.