Everything You Need to Know about Facebook’s Deepfake Crackdown
With all this talk of Zuckerberg’s latest decision to crack down on deepfake videos, it’s time for us as marketers to get to the nitty-gritty of the issue and understand the impact this will have on our work. So, internet users are utilising AI technology to create fake aspects of video media and animating it in various contexts. What’s the big deal, right? Let’s get into the details and find out.
What are deepfake videos?
The word “deepfake” combines the terms “deep learning” and “fake,” and is a form of artificial intelligence used both maliciously and creatively on social media. A deep-learning system can produce a counterfeit by studying photographs and videos of a target person from multiple angles and then mimic their behaviour and speech patterns. Once a fake has been produced, a method known as GANS (generative adversarial networks) looks for flaws in the forgery and improves them to make them more realistic.
Here’s an example (spoiler alert, this isn’t really Salvador Dali):
This technology isn’t new. Hollywood has already made extensive use of it in Solo: A Star Wars Story to insert Harrison Ford’s young face onto Han Solo’s face as well as using it for the acting of Princess Leia in Rogue One. Now, simple software tools, such as FakeApp and DeepFaceLab have made the technology readily available to all, raising concerns about privacy and identity, as well as whether or not it will be possible to replicate even more details of digital identity or attributes.
While AI may provide many exciting advancements in technology, those with malicious intent are developing methods for using these technologies in criminal or ethically questionable ways, most notably in the political arena.
Although the technology offers interesting possibilities and entertainment, the main fear seems to be that it could be used for nefarious ends; “a perfect weapon for purveyors of fake news who want to influence everything from stock prices to elections”, this, according to an MIT technology report.
Deepfakes also raise concern around legal cases as courts often allow footage from victim smartphones or bodycams to be entered into evidence.
The Facebook Shutdown
Deepfakes are a form of spreading misinformation that can be extremely harmful and have adverse effects, as in the case of US politician, Nancy Pelosi.
In this video, the audio was altered to make it seem like Pelosi was drunk and slurring her words.
This is one of the ways deepfakes can be weaponised against celebrities, politicians, religious leaders, influencers et al – anyone is fair game, including Mark Zuckerberg, CEO of Facebook.
This deepfake video shows Zuckerberg gloating about using data to control the future, and is more sophisticated and satirical than Pelosi’s – ironically, making it more believable.
So, it’s no surprise, that after targeting Zuckerberg himself – Facebook decided to crack down on deepfake videos to help curb the spread of misinformation and “fake news” – which the popular social networking site is constantly under fire for enabling.
“How,” you ask?
Well, just as AI can be used to make deepfakes, it can also be used to detect them. Researchers are focusing more on detecting this enhanced media type and finding ways to regulate its use. Similarly, do-gooder technophiles are working on a technique uses Blockchain to verify media sources by analysing their place on the ledger.
But now, there’s a spanner in the works. As with any potentially malicious technology, there are ways in which AI can be used to manipulate media for good. As it turns out, fake media presents us with a grey area.
Deepfake technology is common in popular media
- Will Smith, in his latest movie, Gemini Man, is a good example of the use of deepfake in Hollywood. Here Smith’s 23-year-old clone plays alongside the 51-year-old. The doppelganger is a complete digital recreation, constructed by Weta Digital, a New Zealand-based visual effects studio.
- Ctrl Shift Face, one of the most prolific creators of deepfakes on YouTube, appears in this video seamlessly transforming into Arnold Schwarzenegger Bill Hader impersonates Arnold Schwarzenegger [DeepFake] and then again in this video, transforming into Tom Cruise Bill Hader channels Tom Cruise [DeepFake].
- Another example is BBC One’s drama, The Capture.
- ShamOOK has a composite of various actors, as seen in this YouTube clip, below, while NOVA PBS Official shows how easy it is to be deceived.
How do marketers ethically get around the ban?
Although deepfakes are now banned on Facebook and Instagram, the policy doesn’t extend to the ones meant as parody or satire, where editing has changed the order of words or to “shallowfakes” – a rudimentary form of doctored videos. For these reasons, both the Zuckerberg (satire) and Pelosi (technically it wasn’t a deepfake because it was created without AI) videos will remain online. However, videos will be marked with a warning that it’s false/fake.
Last year Facebook launched the Deep Fake Detection Challenge to create open-source tools that anyone can use to help detect manipulated media for themselves.
According to Vijay Balasubramanian, CEO and co-founder of the fraud prevention firm Pindrop, security experts are bracing for an influx of doctored and misleading videos on social media as the 2020 election in the USA approaches as it is “the perfect catalyst for a flood of these videos,” Balasubramanian told Business Insider. “Any policy to at least say that you’re taking this seriously is a step in the right direction, but I kind of wonder why the [Facebook] policy is so narrowly scoped.”
With this in mind, it seems poignant for other social media sites to follow suit. YouTube has removed the Pelosi video and Twitter has looked into changing their policy around deepfakes but, to date, its policies remain unchanged on this front.
What is the impact of deepfake on marketers and consumers?
With deepfake technology so freely available, influencer marketing will be taken to a whole new level, meaning that trust and transparency will become even more important in any branding/marketing/advertising campaign.
Deepfake videos have been around for about 2 years and are still a relatively new field that is largely unregulated. As with any new technology, they could either have a positive or negative impact on marketers but it’s too soon to predict which route deepfakes will take. In many ways, the power is in your hands as part of the marketing industry. For now, the reactions have been mostly negative but there are a number of positive aspects.
Should you be using deepfake videos in your campaigns?
Let’s weigh up the pros and cons.
- innovative video marketing techniques
- access to new content genres
- more widespread audience, for example, make other demographics more accessible by dubbing into various languages
- entertaining content that will keep an audience engaged
- new job creation in both content creation and legal issues
- potential use in therapy
- personalisation, deeper interconnectivity and emotional interaction
- competitors could create false and misleading content which can severely damage a company’s reputation, with legal complications
- could lead to other types of legal battles such as copyright infringement and slander
- the potential loss of trust from customers, partners and vendors
- some social media sites are already banning the use of these videos
Global use-cases and potential
Some marketing experts have already been quick off the bat to utilise deepfake technology, already providing interesting use cases for how to creatively message consumers.
Malaria must die: A great example of using deepfake video is that of legendary soccer player David Beckham in the Malaria Must Die campaign, creating a petition to end malaria. In it we see Beckham appearing to speak 9 different languages in both masculine and feminine voices. Using AI and deepfake technology to manipulate facial movements, Beckham’s mouth seems to be perfectly in sync with the words. Malaria Must Die campaign. The video was produced by a startup company called Synesthesia in partnership with Ridley Scott Associates for nonprofit Malaria No More.
While this is an extract from her exhibit itself and not an actual ad for the exhibit, conceptually it’s still part advertising. Wearing uses deep fakes in a positive manner and sees it as both relevant and a creative way of marketing her work. Gillian Wearing in collaboration with Wieden + Kennedy 2018. Trailer for Wearing Gillian.
Similarly, we’ve already seen burgeoning potential in various sectors for deepfake technology to make an impact.
Healthcare: The development of deep generative models raises new possibilities that could lead to new and more efficient ways of diagnosing and treating illness. True-to-life data can be generated by researchers as a means of developing and testing diagnosis or monitoring of disease without risking a breach of patient privacy. World Economic Forum – 7 ways artificial intelligence is transforming healthcare. In medical training deepfake images are helping trainee doctors, nurses and surgeons practice their craft.
Film Dubbing, Broadcasting & Entertainment: AI and deepfake technology has the potential for a broad range of positive application, for example, studios can save time and money by employing AI technologies for editing video transcripts after they’ve been filmed and save extra costs of a reshoot.
The Arts: If we refer to our earlier example of the Dali Museum resurrecting Salvador Dali using GANs, we can see a profound use case for using this AI technology in the arts.
Retail: Before long GANs may be used in retail, making it possible for customers to see how they would look in the online products.
All told, deepfake technology could help marketers create innovative campaigns and blur the lines between advertising/retail experiences, and online messaging. It offers brand owners a world of potential within which to ethically interact and conceptualise creative ideas. However, Facebook’s crackdown, as a means towards regulating the spread of fake news, feels like a necessary step towards managing malicious content.
Want to make sure you stay on top of interesting topics and trends in online media? Contact Meltwater for more information on our media intelligence or a free demo.