On Actors and Deepfakes (Part 2)
The second in a series on how media companies can leverage technology to make more money selling content.
Be sure to read Part 1 — On Actors and Deepfakes
Software, powered by the latest in AI and Machine Learning (AI/ML) — Deep Learning neural networks — will soon make deepfakes of any length and perfect quality, on common computer hardware available to almost anyone. That availability is only going to become more prevalent as the tech continues to grow in power and shrink in cost.
You might ask, “what is considered a perfect deepfake?” We can define it as being in-detectable to all humans and most machines.
Digital tools exist today to synthesize images, sound, and video, such as the powerful Adobe suite. Anyone can create “synthetic images” (following terms like fakes, photoshopped, doctored, and touched-up) that are near-perfect quality (mostly imperceptible to a human audience). Many talented technologists use professional CGI software and game engines like UE5 to make video so real you’d think the actors themselves filmed the generated scenes in real life.
It’s not as common yet for the general public to create professional-quality deepfake video or CGI e.g. through video apps like TikTok or on iPhones, due to the powerful computers and servers required. With time this requirement will shrink, thanks to ever-improving computer chips and devices, high-bandwidth Internet, and Cloud services for compute and GPUs. We could soon see perfect deepfakes generated on a common smartphone, putting it easily within reach of both big studios and influencer content creators alike.
So how can the creators of tomorrow benefit from this new technology in the film industry? What aspects can help aspiring artists and filmmakers break in to Hollywood, creating amazing new original content for us all to enjoy?
Positive Usage of Deepfakes
Actors can learn about AI “synthetic media” and how to use it for their career roles, especially the “below-the-line” and aspiring actor trying to break in. Having recently enjoyed Robert Pattinson in The Batman on the big screen, I was excited to see incredible deepfakes of him portrayed in short clips of The Dark Knight. You can see the deepfake below from this amazing creator (Sham00k): https://www.youtube.com/c/Shamook
One way Hollywood studios and actors can benefit is by scaling up sales of existing and future work, by having publishers and distributors translate (localize) for theatrical release in additional international markets like China, Europe, India, South America, Africa, the Middle East, or anywhere else, in any number of languages.
Shoot a TV commercial in English while your product is also available in France? Translate to French and advertise in new markets. Film a movie for the circuits and festivals, and want to show in Cannes? Translate to French, Mandarin, Spanish, and German, letting the audience and judges enjoy watching your work in their native language, bettering your chances of rave reviews, award nominations, and ticket sales.
For now I think there are still many benefits for professional actors to not rely on deepfakes, and we believe there will always be a place for human acting in the world of cinema. It’s important to hone your craft, gain a following and network of references, feel the scenes and roles yourself, perform better through interaction with fellow actors on real locations, with immersive costumes, props, and sounds, which greenscreens simply can’t replicate fully.
Celebrities should also employ a defensive posture, and ensure their management teams are working with Big Tech and the government, to flag unsanctioned deepfakes on the Web, if you can afford it and are popular enough for this to be occurring (and costing you income). Similar to today’s bootleg video, pirated movies, “nudify” deepfakes, and other illegal content, you should ensure your paid content and acting talent is not stolen and monetized by others across YouTube, TikTok, or private streamers.
It sounds scary, but actively providing your likeness data to trusted technical partners, with your library of media content to scan and analyze, will help them in their task searching for and flagging your stolen likeness. Companies like Comcast, YouTube, and Meta (Facebook) are already very good at this, and take down likely millions of infringements every day.
Your Face, Your Right — Actors Taking Ownership of their Likeness
There are nearly 200 countries on the planet, and inevitably some governments will use deepfake technology, while others may try to ban it (possibly even keeping it for their own covert use). Terrorists and autocratic leaders will continue its use and soon likely cause a serious international incident, easily attacking overs-seas with basic 2G connections. Deepfakes will be generated on even out-dated yet powerful smartphones, on some future illicit app anyone can put together with open-source algorithms and model weights.
Unfortunately, some day we may also see illegal datasets of human faces and videos on the dark web for rogue actors to train ML Models with. We see this today with hacked company data that is regularly stolen to be sold there. The dark web is also known to hold illegal media, “child porn” and “revenge porn”, some stolen and some generated. It is hard to control, and may become important for worldwide governing associations like the UN (UNICRI), NATO, EU, WHO, Interpol, NIST, OECD, or others clamping down on the spread of harmful deepfakes and any illegal data models.
We hear people on a regular basis across SF and LA tell us that they hate AI and “smart robots”, and hoped humanity wouldn’t create this technology. To an extent I agree in that I recognize the concern around AGI. I certainly have a healthy respect and fear for AI, and am aware of ways it can be used for harmful purposes. It’s critically important companies think about AI and study the basics through educational content like this free course: https://www.coursera.org/learn/ai-for-everyone
It’s never smart to ignore technology and hope it goes away. Use it to your advantage, and decide when to draw the line and limit usage especially with respect to rights and privileges around your likeness (look and voice).
Click below for more articles from Russell on AI and Filmmaking.
If you enjoyed reading, please share with your network and subscribe! Here on Medium, Discord, Reddit, or other fine communities.
References
Top image from Entertainment Weekly (EW)
Middle photo from Pexels (Photo by Tima Miroshnichenko)
Bottom photo purchased from iStock
Robert Pattinson deepfake (Sham00k created using DeepFaceLabs) https://www.youtube.com/watch?v=dmuYz0aZGgU
Mark Hamill deepfake (Sham00k created using DeepFaceLabs) for Star Wars episode (spoiler alert) See also: https://www.creativebloq.com/news/star-wars-deepfake
https://www.youtube.com/c/Shamook
AI NGOs
Other active AI NGOs include MIRI, ACM, IEEE, Stanford University HAI, OpenAI, Montreal AI Ethics Inst., DARPA, Future of Life Inst., AJL, CIS, CHAI, CSER, CITP.
Free-ish Courses (MOOCs) AI/ML
[1] AI For Everyone (Andrew Ng) Offered by DeepLearning.AI on Coursera.org (link)
[2] Machine Learning for Business with Python — CS 68 (Charlie Flanagan) Offered by Stanford Continuing Studies (link)
[3] Machine Learning (Andrew Ng) Offered by Stanford on Coursera.org (link)
[4] Prediction Machines — The Simple Economics of Artificial Intelligence (Agrawal, Gans, Goldfarb) HBR Press 2018 (book)
More books to read about AI Safety
[1] Superintelligence — Paths, Dangers, Strategies (Bostrom) Oxford University Press 2014
[2] The Age of AI — And Our Human Future (Kissinger, Schmidt, Huttenlocher) Little, Brown, and Co 2021
[3] Our Final Invention — Artificial Intelligence and the End of the Human Era (Barrat) Thomas Dune Books 2013
[4] The Singularity is Near — When Humans Transcend Biology (Kurzweil) Penguin Books 2005
[5] AI Superpowers — China, Silicon Valley, and the New World Order (KF Lee) Houghton Mifflin Harcourt 2018
🤖🎥
Copyright © 2022 CYBERFILM.AI CORPORATION
