5
min read
Cognition and Creativity
Risks of AI

Deepfake

March 13, 2024

Defining deepfake

Deepfake is a term used to describe the creation of fake artefacts, usually people, based on learning from images and speech of the real person.

How it's used

Various well publicesed examples have been around for a few years, such as the Channel 4 production of a fake speech by the UK’s Queen, broadcast at Chritmastime in 2020. Various politicians and celebrities from Barak Obama, Mark Zuckerburg to Tom Cruise have also suffered from being used in deepfake videos (Creative Bloq). The algorithms are also being used to bring people “back from the dead” by using old recordings and images to create interactive avatars of people who have died.

AI used to resurrect Salvador Dali. Perhaps the most famous example of deepfake is a recreation of Salvador Dali from old film clips and speeches of Dali. It is used as a host visitors can interact with at the Dali Museum in Florida.

Deepfake is perhaps best known for its use in non consensual pornography, where images of a person can be used to superimpose upon a pornagraphic scene, making it look as though that person is involved in a sexual act.

Film Production

The same technology however is seen as a gift to film producers, usually referred to as AI generated videos or synthetic media, enabling them to cut down on editing time or even remaster old videos.

Targeting children

More recently, since free algorithms have been made available, even children are now the targets of bullying from others when pictures of them posted on facebook are manipulated by AI to appear on someone else’s body, even Dustin Hoffman’s autistic character from Rain Man (iNews).

Fake or real - can you tell?

In the past, it has been fairly easy to spot deepfake images, but now the technology is becoming so realistic that it is often almost impossible to distinguish between a deepfake and the person targeted. The usual area where deepfake can be spotted most readily is in poor lip sync.

Dangers of deepfake

Over 1.2 trillion digital images and videos were captured in 2017 – a figure that increases by about 10% each year. Around 85% of those images are captured using smartphones, carried by over 2.7 billion people around the world.

Deepfake has huge implications for everyone, not just politicians or celebrities, but also school children and anyone whom someone has a grudge against. The opportunities for criminals to blackmail, corrupt states or actors to manipulate news and politics to destabilise are obvious.

As we all know, mud sticks, so even if the images or video, is proven to be deepfake, the reputational damage has been done. The use of deepfake in these ways is nothing less than identity theft.

Possible solutions

The only way to protect against these challenges is to encode all digital data using blockchain technology so that an audit trail of where the data has been and how it has been used is possible. This, together with a recognition, internationally, that data associated with a person belongs to them, akin to Copyright or Intellectual Property, would provide some deterrent and at least the means to uncover deepfake and to prosecute bad actors.

An example of how blockchain is being used to combat fakenews is Safe.Press, a news certification service developed by Block Expert in France. The Content Authority Initiative with partners like Adobe, Twitter and the BBC is another venture seeking to address the problem using technology developed by Start up company, TruePic. TruePic uses blockchain technology currently but is developing a more scalable system using public/private keys based on a Public Key Infrastructure.

The DeepTrust Alliance is a not for profit initiative seeking to bring together a global coalition of stakeholders joining the fight against digital disinformation and deepfakes. As they state on their website -“This is an arms race. Solutions require human and technical interventions to counter threats — and we’re in a race to out-innovate nefarious actors.” 

References

Michael Grothaus, Trust No One: A Journey Into Deepfakes, Hachette UK,  2021. [https://inews.co.uk/news/technology/deepfake-videos-school-bullying-cyberbullying-ai-apps-parents-teachers-children-1290664]

Creative Bloq [https://www.creativebloq.com/features/deepfake-examples]

Salvador Dali [https://youtu.be/mPtcU9VmIIE]

https://www.deeptrustalliance.org/

https://www.weforum.org/agenda/2020/03/how-to-make-better-decisions-in-the-deepfake-era/

G