DeepFakes: The Good and The Bad

DeepFakes: The Good and The Bad

What Are Deep Fakes?

In one sentence? Really good fake data. 

If you have seen the video where Mark Zuckerberg had claimed about having total control of billions of people’s stolen data, you have seen a DeepFake. Deepfakes are media in which a person in an existing image or video is replaced with someone else’s attributes. The term itself comes from the technology that powers it, i.e, Deep Learning. It is a subset of Artificial Intelligence that has networks capable of learning from unstructured data.

How are they made anyways?

Using Deep Learning – a subset of AI – as we mentioned earlier. But, let us give you some context first. 

There’s a type of deep learning AI called Generative Adversarial Network or GAN which is called the perfect cat-and-mouse game. Let us explain. There are two sides of the neural network – the Manipulator and the Enforcer. 

Here, Manipulator – the part that tries to create a new or fake video, image, audio using its training data. And Enforcer – tries to detect the fake output. At first, the manipulator can only create terrible fakes but as it learns (Machine Learning), it gets better. Once the Manipulator gets better, the enforcer has to get better if it is to spot the fakes. And in this game of cat and mouse – they both become better and better. 

This is the technology used to create a Deepfake. We give GAN some rules and it creates the perfect fake for us. For example, this site thispersondoesnotexist, churns out images of people who do not exist – but are created using AI. 

Not a real person.
Also not a real person.

Is Deep fake good or bad?

Walter Benjamin, a German cultural critic who lived in the early 1900s, wrote:

We don’t need to fear new forms of art that are augmented by technology as long as we have a shared understanding that good governance will prevail.”

This is exactly what it is about. There are moral concerns surrounding every human using any sort of technology – the technology anyway, at the end of the day – can’t itself do the ‘right’ or ‘wrong’ – right? 

The widespread concern about Deepfakes is its potential to spread misinformation and discredit individuals. We have been using video, photo and audio as evidence in our civil and criminal justice systems but what is the next step if these could be manipulated easily and perfectly? Will we, as a society, have to move on from relying on our sense of sight and observation to establish objective facts? While there are ways through which deepfakes can be detected, to realise it is a case of deepfakes might take us some time.

So is Deepfake a good thing that has happened to us or bad? Depends mostly on how we use it. Let’s see how it has been used until now –

The Good:

  1. Deepfakes accelerating accessibility:

Media created using Deepfakes are making accessibility tools smarter, personalisable and affordable. Microsoft’s Seeing.ai and Google’s Lookout are using AI for recognition and creating deepfake voices to narrate objects, people, and the world for the visually impaired.

  • CaneTroller, a haptic cane controller, has finally made the VR world inclusive. 

V9, a participant testing out CaneTroller at Microsoft said:

“It was exciting to … be able to do [VR] as a blind person using [Canetroller]… it was inclusive. I enjoyed getting to participate in something… that sighted people get to do.”

  • VOCALiD uses voice blending technology to create unique vocal personas for any device that turns text into speech for those with speech and hearing difficulties.

Rupal Patel, CEO of VocalID:

“VocalID is now on generation four of our voice synthesis engine, which allows us to make the most highly replicable voice of an individual. [While we focus] mostly on individuals who are nonspeaking, [we also help] individuals who were about to lose their voice … Prior to them losing their voice, they bank their voice [with us] and we recreate it.”

2. In the Education Industry:

  • Deepfakes are bringing historical figures back to life for a more engaging and interactive classroom. A synthetic video of historical figures has proved to be a better learning tool and experience in recent days. For example, JFK’s resolution to end the cold war speech, which was never delivered, was recreated using his synthetic voice and speaking style:
  • Human anatomy, industrial machinery, and complex industrial projects are being simulated in a mixed reality world to teach students through Microsoft’s Hololens.
  • Udacity, an online learning platform with over 100,000 courses, used Deepfakes to create video lectures automatically for its MOOC content.

3. In the Entertainment Industry:

Deepfake – if realised to its full potential, may be able to bring a completely new era for the Entertainment Industry. It is already being used by YouTube creators like ctrl shift face and shamook to create movies or give existing ones a new look.

  • Deepfake technology can democratize costly VFX technology as a powerful tool for independent storytellers at just a fraction of the actual cost. 
  • Dalí Museum in St. Petersburg, Florida, recently created an exhibition called Dalí lives, bringing Salvador Dali, the famous Spanish surrealist artist renowned for his technical skill and precise draftsmanship, back to life using deepfakes for visitors to interact and take a selfie with.
  • Samsung’s AI lab in Moscow brought Mona Lisa to life by using Deepfake technology.

4. Gaming Industry:

Nvidia recently demoed a hybrid gaming environment created by deepfakes and is working on bringing it to market soon.

  • Deepfake voice is being used to narrate storybooks. The author’s synthetic voice font can be turned into the author’s book’s audio format. Businesses are starting to use synthetic voice-overs of the same actor in different languages to broaden the reach of their content. 

5. Deepfakes as an empathy tool for the online world:

  • Deep Empathy, a UNICEF and MIT project, simulates how cities around the world would look amid a conflict similar to the Syrian conflicts. The project created synthetic war-torn images of Boston, London and other key cities around the world to help increase empathy for victims of a disaster region.
  • Several voice-technology startups are creating Deepfake voice as a bereavement therapy or help people remember the deceased and connect with them.

6. In-Reach and Message Amplification:

A health charity in the UK partnered with David Beckham and an AI company recently to produce a video and voice campaign to help end malaria. Beckham spoke in nine languages seamlessly done using the deepfake technology. 

7. Deepfakes are taking influencer marketing to another level:

  • A politician in India created a deepfake to reach out to different linguistic voters just a few days before the Delhi elections. It is the first documented use of a deepfake by a political party for campaigning purposes.
  • AI Foundation, a company that builds personalised AI tools, is creating personal AI for influencers and celebrities to help them engage and amplify their reach with their audience. They recently launched a personal digital avatar of Deepak Chopra.

8. Public Safety

  • A team of civil investigators used cell phone videos, autopsy reports, and surveillance footage to reconstruct a virtual crime scene using the Deepfake technology in Ukraine. Read more about it here.

9. Sports

  • Reuters recently demonstrated a fully AI-Generated deepfake presenter-led sports news summary system to help personalise news at scale.

10. Retail

  • Deepfakes are turning customers into models by helping them virtually try out the latest apparel and accessories in the fashion retail business.
  • Data Grid, a Japanese artificial intelligence company, created a Deepfake AI engine that automatically generates virtual models for advertising and fashion.

The Bad:

1. Threat to Individuals:

The neural network AI is being used to make a video of any individual lip-sync to the audio input of one’s choice. This can and has potentially harmed people’s reputation. Politicians and celebrities are the most common victims of deepfakes. 

  • For example, in this Instagram video posted by @bill_posters_uk, a deepfake of Mark Zuckerberg talked about how he had control of billions of people’s stolen data.
  • When Donald Trump Joined Breaking Bad: Donald Trump’s face was merged with Bob Odenkirk’s face to recreate this scene:
  • Targeting Women:

Deepfakes are being used to create pornographic videos causing emotional and reputational harm towards individuals, mostly women. According to the Deeptrace, now sensity.ai report, 96% of deepfakes are pornographic videos, with over 135 million views on the pornographic website alone.

Danielle Citron, a professor of law at Boston University: 

“Deepfake technology is being weaponized against women.”

2. Threat to Society and Democracy:

Deepfake of a political candidate can sabotage the reputation of the candidate and may alter the course of that election.

A Belgian political party created a fake video of the U.S. president in which Mr Trump calls Belgium to exit the Paris climate agreement.

3. Threat to Businesses:

Deepfakes are being used to impersonate business leaders and executives and facilitate fraud:

  • A CEO of a UK based energy firm was asked to wire $243,000 to a Hungarian supplier by his boss over a phone call. Only once he had transferred the money did they realise that it wasn’t his boss whose voice he heard but that of a Deepfake manipulated voice. 
  • According to Symantec, millions of dollars were stolen from three companies, which fell victim to deepfake audio attacks. They each received a call from their CEO [Deepfake] requesting an urgent money transfer on each attack. The deepfakes models were trained on the CEO’s public speeches. 
  • Just a few months back, a Pennsylvania attorney was fooled by a deepfake voice of his son, claiming that he needed $9,000 as bail money.
  • A businessman was tricked by a bunch of Israeli fraudsters after they impersonated the French foreign minister in a Skype video call and asked for $9 million.

Just like any other technology that we use, there can be a day where Deepfake’s potential to do good will be realised and integrated into our lives. The question is: will we ever stop misusing ingenious technologies? We wrote this blog to bring forward the case of Deepfakes: the ethics aren’t black and white.

Key Takeaways:

  1. Deepfakes are media in which a person in an existing image or video is replaced with someone else’s attributes.
  2. The widespread concern about Deepfakes is its potential to spread misinformation and discredit individuals.
  3. Many companies are finding out innovative ways to use Deepfakes for the good of humanity.

Resources:

If you have any interesting ideas around DeepFakes or AI in general, let’s have a conversation. Feel free to reach out to us at info@aidetic.in.

Leave a Reply