Disclaimer - this post has been updated in September 2024 as a YouTube video has been made about this!
Welcome, dear readers, fellow humanoids, and fellow robots (if you're reading this or watching my video), to a rollercoaster ride through the crazy-a** world of deep fakes. Get your seat belts ready, because we are about to embark on a WILD journey that's both mind-bending and side-splitting. From celebrity impersonations such as Leo and Margot to Tom Cruise on TikTok to political shenanigans, deep fakes have taken the internet by storm, leaving trails of laughter, confusion, and that ever-growing existential dread in their wake. Yep, back to the existential crisis we go!
Something that I've wanted to learn a lot more about is AI and I wrote that in my "30 things I want to do before I'm 30" blog post last year, and because AI is such a trending topic right now, I thought why not? I also really didn't know what I should discuss on the blog next so giving myself a challenge was something I wanted to do. And I want to say, I'm not an expert in anything AI-related, but I would say I'm a bit of an expert Social Scientist - I read people (and no not in that psychic way, I don't know what you're thinking). As creepy as it sounds, I just watch and pay attention to the quirks and ways people behave and that gives me a bit of a good read. I would say studying both Sociology and Psychology really helped me to be able to do this. I would say it's a bit like that show "The Mentalist" but not as good as Simon Baker's character Patrick was. And, to toot my own horn here, I did write a whole masters thesis on Social Media - Social Media: A New Age Religion? to be exact. And in saying that, I feel like AI and social media are kinda blurring together, becoming "hyper-real" if you will. For example, think of all those celebs like Katy Perry who everyone assumed went to the Met Gala but they weren't actually there. Did you fall for it the same way Katy Perry's mum did? I saw the photo after it was announced it was AI, but the heads in the background of the photo kinda gave it away.
Anyway, what's a deep fake Ashy? According to Wikipedia, deep fakes "are synthetic media that have been digitally manipulated to replace one person's likeness convincingly with that of another. It can also refer to computer-generated images of human subjects that do not exist in real life." The wiki article goes on to say that creating "fake content" isn't ALL that new in 2024 but the leveraging of AI and machine learning is what's really freaking people out. I've had people tell me how bad Tik Tok is with some of these deep fakes and sure, Tik Tok is allowing it but don't forget all the other social media platforms and trustworthy news sources who allow this type of content too. My personal opinion is that it just needs to have a disclaimer i.e., don't say it's all user-generated content if it's made by AI, just say it's a deep fake (and you have permission from the person you are imitating). There's also a whole other debate on censorship when it comes to this. I'm all for free speech as long as it's not harming anyone - I hate that subreddit 50/50. I won't go into detail with what that's about but you get the gist of where I'm going.
I also think it's important to note what GANs are (Generative Adversarial Networks). Basically, it's a type of machine learning that uses two models to compete (battle of the machines lol) against each other until the second one is unable to distinguish between a fake video and a real video. This means that a deepfake would appear realistic to human viewors as we'll see below. However, creating deep fakes requires a lot of data, hence why deep fake creators may target politicians or notable celebrities.
I wanted to show some examples of the deep fakes online that I mentioned earlier like the Margot Robbie and Leonardo Dicaprio ones.
Unfortunately, I can't embed Tik Tok videos on my website (hopefully that feature comes soon) so I'll just have to provide links and screenshots.
Margot Robbie and Leonardo DiCaprio Deep Fakes - https://www.tiktok.com/@fake_dicaprio/video/7361503407036763422
I feel like Margot is a little bit more convincing than Leo is but I also want to give a major shout out to the creator of this - the username says it's fake and also puts it in the video so you're aware. Leo, on the other hand, is giving a little bit of uncanny valley vibes in this video.
Moving on. Tom Cruise Deep Fake - https://www.tiktok.com/@deeptomcruise/video/7039410659607661870?lang=en
Again, respect to the creators for making it super obvious it's a deep fake but this one has always got me feeling a little off. I don't know if it's because it looks and sounds exactly like Tom Cruise or if I just find Tom Cruise a bit creepy in general. I mean, isn't he at the top of Scientology or something? Maybe that's what it is, or maybe both of them are just creepy as.
If you're wondering how far it goes with "everyday people" - check out this video from the BBC UK. It's pretty freaky how far people go. TW: s*xual content.
I feel awful for this woman. The things that she endured were absolutely horrible. But I can also relate in a way because I once had my social media accounts and emails including my old Reddit account hacked by someone I trusted. If anything, it taught me not to trust people. If you want to know about that, let me know in the comments below and I'll dedicate a blog post to a story time. It's nowhere near the same, but it is that scary feeling of having someone you trust break your trust.
I also want to show some of the images from the Met Gala for reference!
Katy Perry AI generated image (look at the heads) - image from "here".
From left: Rihanna, Katy Perry's "other" look, and Lady Gaga. Image from "here."
I feel like whoever made these, just wanted to capitalise on the Met Gala. Maybe they have an agenda or maybe they're just trolling. Or maybe, there's more to it against women and children? "This article" takes a look at why that might be.
Now that we've had a look at some examples, I want to look at the pros and cons of Deep Fakes. The world's getting scary isn't it? Black Mirror has never felt so real.
The Pros of Deep Fakes
Entertainment Extravaganza: Let's face it, deep fakes ARE actually entertaining. They're creepy for sure, but they might give someone a laugh, or make someone enjoy some random internet content when they're feeling down. Comedy and social media/UGC can take on new heights with deep fakes. Options are endless.
Creative Possibilities: Although sometimes comical, deep fakes aren't just for laughs, they do have some creative benefits to them as well. Artists and filmmakers can use this technology to explore new storytelling techniques, from digitally resurrecting long-deceased actors to creating immersive virtual worlds that blur the line between reality and fiction. There's that "hyper-reality" again.
Educational Value: Believe it or not, deep fakes have educational potential. By demonstrating the power of AI and machine learning algorithms, they're sparking important conversations about media literacy, digital manipulation, and the importance of critical thinking in the digital age.
Marketing/Hyper-personalisation: Deep fakes can help business engage in more personalised advertising to encourage users to become consumers. Aren't we already consumers though? Yep, we consume everything. In this article, "Bartoletti said, “deepfakes will likely be used in business as avatars in presentations, video conferencing, and metaverses. Such avatars could be used to mask identities and avoid bad-hair days. Deepfakes could also allow personalisation of news and sports feeds and modeling clothes by mere mortals.” Bartoletti is the privacy and data protection expert at IT consultancy Wipro.
Low-Cost Video Campaigns: I kinda think this one is self-explanatory but basically a company doesn't need to higher a really expensive actor for a campaign if they can just fake one. That way, spending less on campaigning might mean a higher turnaround revenue for that year. It is a bit more complicated than just faking it though - they'd need to buy the actor's identity license.
The Cons of Deep Fakes
Reality Check: As entertaining as deep fakes may be, they also pose a serious threat to our perception of reality. With the ability to create convincing videos of public figures saying and doing things they never actually did, deep fakes have the potential to sow chaos and confusion on a global scale. What was real? What's not real? It could cause a major storm in the political world, especially with those with a lot of power. For example, the robocalls using AI to impersonate calls from Joe Biden. You can read about that "here".
Privacy Predicaments: Deep fakes raise significant concerns about privacy and consent. Anyone with access to enough photos or videos of an individual can create a convincing fake that could be used for nefarious purposes, from revenge p*rn to political smear campaigns. See the BBC video I linked above for one case.
Ethical Quagmires: The ethical implications of deep fakes are vast and thorny. From impersonating celebrities for profit to manipulating public opinion with fake news, the possibilities for abuse are endless. As the technology continues to advance, people in society must grapple with tough questions about regulation, accountability, and the boundaries of free speech. Where do we draw the line?
Exploitation: Deep fake technology has the potential to exploit medical data which might lead to an incorrect diagnosis. See more information "here".
Spread Misinformation + Propaganda: We all know what "confirmation bias" is, I assume, well I hope so, where if we go looking for something, we're gonna find the answer we want. So we can kinda understand why some people are more likely to believe in conspiracy theories than others. Or maybe that's just my opinion. Anyway, the point is deep fakes have the potential to spread even more misinformation and propaganda online. Here's a quote that sums it up really well:
The ability to create convincing fake videos of political leaders can lead to confusion, disinformation, and a loss of trust in public institutions. In addition, such forms of disinformation could manipulate and distort the complicated media ecosystem.
Taken from "this source".
So basically, deep fake technology has both pros and cons. I don't know whether the pros outweigh the cons, because the cons are really scary. But, I do know that I find it interesting at the least. Personally, it could be all about the "intent" - are deep fakes intending on harming society? Or are they just there for entertainment purposes? Are deep fake creators just trying to create art? All questions to ponder.
Face-swapping technology can also be seen in other applications such as Snapchat (think about the filters you're using next time you put one on before you go to snap a selfie) and FaceApp (a face editor). I feel like this issue stems further than just to do with technology, it can even relate to how we perceive ourselves - like are beauty standards going to change because of deep fake technology? Women are already encouarged to lose weight, gain weight, have a skinny waist, but have bigger b**bs. Men are encouraged to be muscly, fit, and sporty. I feel like there's a connection between beauty standards and deep fakes, but it's just not super obvious yet.
So what are some tips for spotting a deep fake or an Al altered image?
Look for inconsistencies, if something feels "off", it probably is. Lighting might be not what you expect and reflections might not be where they usually are.
Check the audio quality - in a fake video, audio probably won't be the same as a real video, so make sure you're listening.
Look for unnatural movements - if you can't move like they are, then it's probably fake (OR they're just really talentend).
Compare and contrast with known sources.
Ask someone in the know.
And if you're not sure, assume it's fake. Remember that time a fake image won the wildlife photography competition? One of the wildlife comps but I can't find the photo or any news articles. I will update once I've found it. Ahh yay, I found it "here". It wasn't AI specifically, but the photographer had used a taxidermy animal and passed it off as real.
In conclusion, as we emerge from the murky depths of our deep dive into deep fakes, one thing becomes abundantly clear - while they may be a source of endless entertainment and creative inspiration, deep fakes also come with a hefty dose of risk and responsibility. As consumers of digital media, it's up to us to approach deep fakes with a critical eye and a healthy sense of skepticism. So the next time you encounter a video that seems too good (or too absurd) to be true, remember to take a step back, do your research, and above all, don't believe everything you see. Happy deep-faking, fellow truth-seekers!
Some readings:
Gambín, Á.F., Yazidi, A., Vasilakos, A., Haugerud, H. and Djenouri, Y., 2024. Deepfakes: current and future trends. Artificial Intelligence Review, 57 (3), p.64.
Mirsky, Y. and Lee, W., 2021. The creation and detection of deepfakes: A survey. ACM computing surveys (CSUR), 54 (1), pp.1-41.
Masood, M., Nawaz, M., Malik, K.M., Javed, A., Irtaza, A. and Malik, H., 2023. Deepfakes generation and detection: State-of-the-art, open challenges, countermeasures, and way forward. Applied intelligence, 53 (4), pp.3974-4026.
Sample, I 2020, 'What are deepfakes - and how can you spot them?' The Guardian - https://www.theguardian.com/technology/2020/jan/13/what-are-deepfakes-and-how-can-you-spot-them
Verdoliva, L., 2020. Media forensics and deepfakes: an overview. IEEE Journal of Selected Topics in Signal Processing, 14 (5), pp.910-932.
Keep fakes is a problem that's getting quickly out of hand. Great insight. Great post!