Deepfakes — A Disaster for Trust or Business Opportunity?
There is little doubt that you by now have heard about deepfakes or at least seen one on Youtube or elsewhere online. They’ve been written about by almost every major media publication and deepfake videos online (most of them satirical, PSA’s about the dangers of deepfakes, or if you venture into the darker corners of the internet, pornographic) have amassed hundreds of millions of views.
If you were to do a quick Google search of ‘deepfakes’ the vast majority of articles that would surface would be about the dangers of deepfakes, how they signal the end of truth as we know it and the more sordid use cases of deepfake technology ranging from political manipulation to revenge porn. It is true that deepfakes have so far mainly been used for nefarious purposes. A 2019 report from Deeptrace estimated that 96% of deepfake videos that they were able to find online were pornographic in nature. In the realm of politics, most deepfakes have been in the form of PSA’s warning about the dangers of deepfake technologies, such as a Buzzfeed’s widely viewed Obama deepfake video using comedian Jordan Peele to create a warning about the usage of the technology and to verify outrageous or surprising claims before holding them as the truth, even if it’s in the form of a video. After all, once deepfakes get good enough, then seeing may no longer be believing.
The threat of deepfakes shouldn’t be downplayed. Deepfake technology can be incredibly dangerous if used in nefarious ways, in pornography, it will most likely disproportionately harm women who are already affected by revenge porn, and it will no doubt be weaponized by bad actors who want to propagate misinformation in order to sow division in other countries, discredit journalists and others, used to try to make a profit, for fraud, or just used because some people want to watch the world burn. But, because of the work of companies like Deeptrace and Sentinel (full disclosure, I am an advisor to Sentinel) who are building deepfake detection technologies and the excellent work by a myriad of media organizations and nonprofits who are educating individuals on how to spot deepfakes and about digital literacy more broadly, there is hope that this technology won’t be as destructive as it could be if no one was working to stop its negative externalities.
However, very little examination has been given to the potential benefits of deepfake technology and the business opportunities that it unlocks. The technology underlying deepfakes is inherently neither good nor bad and can be used by entrepreneurs for either. Like many technologies, deepfake applications were picked up by the government and the porn industry first, but there has already been an emergence of applications for deepfakes in the broader entertainment and media industries, in the translation field, and in marketing and advertising.
To find out what the newest potential business applications of deepfakes were, we reached out to the folks at Tribe.ai, a collective of some of the top AI-focused engineers from places like Facebook, OpenAI, and MIT who come together to work on interesting and pressing projects. Unsurprisingly, we received a myriad of interesting ideas, some pretty out there. Below are the top three ideas (lightly edited for clarity).
- Personalized News: Based on a viewers' demographic or favorite newscaster, a deepfake could be used to “fake” the news as it is being reported so that you could always watch the news as reported from your favorite anchor or newsperson. Of course, this isn’t limited to just other newscasters, you could have Chris Pratt and Jennifer Lawrence report on the latest updates on the stock market.
- Simplified Animation: Where most hyper-realistic CGI such as the work done by Disney to input Carrie Fisher into the latest Star Wars movies is extremely time and cost-intensive, the use of GANs could help future animators with the generation of variants on textures, objects, and patterns and dramatically bring the total time and cost down for animation.
- Remote Work: Becoming extremely topical with the spread of COVID-19 and lockdowns on travel and even small meetings across the world, being able to work remotely is critical. For the entertainment industry, which is no doubt in overdrive given how many people (both adults and children) are locked indoors, there is an urgency to produce content but a danger to doing so in person that most aren’t willing to take — as evidenced by the cancellation of SNL’s current season among others. However, using GANs, actors could record themselves remotely and the final video could be stitched together (note: this won’t work for scenes with heavy movement, but for a show like Last Week Tonight or most other late-night TV shows, this would work exceedingly well).
As mentioned above, there have been uses of deepfakes in a variety of other industries as well. Some of the most notable examples are Synesthesia’s work creating a deepfake of David Beckham speaking multiple languages to raise money for charity, CannyAI’s work across a range of fields with localization projects done for companies like Accenture, satirical videos produced for a range of audiences, reshooting for movie production, and much more, Rosebud AI’s AI-generated models, Respeecher’s voice cloning technology for dubbing, audiobooks, and more, and finally, the work of Paul Shales (The Fakening) who has created dozens of deepfake videos including for artists like Diplo and The Strokes.
Like most questions in the real world, the answer as to whether deepfakes will be a net negative for society or whether there are useful and non-destructive applications that outweigh the bad is a loaded question with no simple answer. Although most of the current attention has gone to the negative aspects of deepfakes and they have the potential to be incredibly dangerous for society, there are people working to identify deepfakes and make sure that malicious actors are stopped before they can cause serious damage and there are a myriad of emerging use cases for the technology that could help to revolutionize the entertainment industry, create meaningful advances in marketing personalization, and more.
