Tackling Deepfakes

What are Deepfakes?

Deepfakes are videos, images and audios created or edited with a form of artificial intelligence (AI) model to create a convincingly true version of a real video/image or audio.

How are deep fakes created?

It is created using techniques in machine learning (ML), a subfield of AI especially Generative Adversarial Networks (GANs)

In the GAN process, two ML systems called neural networks are trained in competition with each other. 

1. The first network, or the generator- tasked with creating counterfeit data such as photos, audio recordings, or video footage that replicate the properties of the original data set. 
2. The second network or the discriminator, is tasked with identifying the counterfeit data. 

The generator network adjusts to create increasingly realistic data based on numerous iterations fed by the counterfeit data until the generator improves its performance such that the discriminator can no longer distinguish between real and counterfeit data.

Potential uses of Deepfakes technology:

1. Medical purposes

GANs can be used to synthesize fake medical images to train disease detection algorithms for rare diseases and to minimize patient privacy concerns.

2. Entertainment purposes
   1. It can accelerate the speed of game creation.
   2. Can mimic artists in audio, video or image forms including those who have deceased.
3. Education
   1. An educator can use deepfakes to deliver innovative lessons that are far more engaging than traditional visual and media formats.
   2. For instance, it can bring historical figures back to life for a more engaging and interactive classroom.
   3. Synthetic human anatomy, sophisticated industrial machinery and complex industrial projects could be modelled and simulated in a mixed reality world to teach students in an interactive manner.
4. Autonomy in speech and expression
   1. Deepfakes can help human rights activists and journalists to remain anonymous in dictatorial and oppressive regimes.

Concerns associated with Deepfakes:

1. Affecting the credibility of elections
   1. Deepfakes can be used to spread disinformation resulting in erosion of public trust and creation of chaos.
2. Fake-driven financial frauds. For instance, voice can be used to mimic individuals for financial frauds or break into the security systems of enterprises.
3. The data stored in cloud servers pose a serious threat as fraudsters can use the data of images and videos for AI models in creation of Deepfakes.
4. Leads to erosion of trust in video and images irrespective of its authenticity.
   1. This could also lead to negation of reality and acceptance of fake contents.
   2. Politicians can also claim in future that solid evidences as doctored when caught red handed.
5. Elderly population are more vulnerable as they have poor knowledge about deepfakes content.
6. Difficulty in discriminating the real content and fake content as it needs huge computing power and forensic auditing.

Way forward:

1. US Big-Tech companies have voluntarily agreed to watermark AI-generated content to ensure safety.
2. The European Commission has made it mandatory to label AI generated deepfakes under the Digital Services Act, non-compliance attracting huge penalties.
3. Developing a Code of Conduct by the AI companies for ensuring productive applications of deepfakes.
4. Awareness among public especially vulnerable sections has to be created to protect themselves from the deepfakes bait.

Link: With deepfakes getting better and more alarming, seeing is no longer believing | The Indian Express