Digital avatars or deepfakes ?

Kongbrailatpam Rajeshwar Sharma *

With the rapid strides in information technology, deep learning algorithms and generative artificial intelligence or AI, one can easily transfer money or pay with a click of a mouse-button or a touch of screen within seconds otherwise it would have taken a number of hours or even days in the not so distant past.

On the other hand, the technology of generative AI and deep learning algorithms has the potential to create deepfakes and misinform the people, particularly the naïve and trusting netizens. Not only can one send text messages, fake images and videos generated by AI, but one can also stream live deepfaked videos online.

Last November, a live video of a digital avatar of a dead woman who was introduced as Duwarka, daughter of Vellupillai Prabhakaran the chief of the erstwhile Tamil Tigers, was shown giving a speech on YouTube. She had been killed in an airstrike in 2009 during the fag-end of the Sri Lankan civil war.

So advanced is the technology behind generative artificial intelligence that fake images, videos and audio of any person can be generated with the technology. Such fake or unverified images and videos are constantly posted online news portals and on many social media platforms perhaps with an ulterior motive.

In other words, lies or misinformation are being spread day in day out on several social media platforms using generative AI and deep learning algorithms all over the world. In March 2023 AI generated images of Donald Trump, former President of the US, being arrested were circulated online.

They looked so real that some people shared the images on social media. There is no dearth of AI tools or application software that can generate fake images and videos with the potential to incite violence. To name a few of them, Midjourney, Dall-e and Stable Diffusion are generative AI tools which are used to create deepfakes. They enable a user to swap faces and generate realistic images and videos in a few minutes. They are easily available on the Internet.

Midjourney is a generative adversarial network or GAN AI. It has been developed to generate high quality images of people, objects or even landscapes that look realistic and authentic. Dall-e is another GAN AI which was named after the surrealist artist Salvador Dali and the movie character Wall-E. This AI algorithm is designed to learn from a large set of image data and generate a wide range of images based on textual commands.

Stable Diffusion is a powerful AI algorithm or AI tool that can produce realistic images and videos. One of the key features of Stable Diffusion is its ability to transform a state of image into a different state.

It can make a person, whose eyes are closed in the image, open his eyes in the newly generated image. Stable Diffusion can make a person look happy or sad. Bhashini is another generative AI tool developed by the government of India. It can translate a speech given in Hindi into Tamil or any other Indian language in real time. Bhashini was widely used during the Lok Sabha election campaign recently.

In the eyes of law, deepfake is nothing less than forgery. In other words, it is an act of digital forgery. Artificial intelligence and machine learning are used in creating deepfakes to produce images and videos that appear to be realistic and authentic. The process involves, first of all, creation of media files by superimposing a person’s face on another’s body or by altering his or her voice or facial expressions and body movements in a video.

Nowadays social media platforms are riddled with deepfakes because they can be created with open-source software or customized tools. Recently two deepfaked videos of Bollywood stars Ranveer Singh and Aamir Khan went viral. They were shown campaigning for the Congress Party. The two viral videos are classic examples of deepfakes.

Another viral video showing heavy rainfall and flood in Dubai was a deepfake that fooled the world. On 30 April this year, two people were arrested in Delhi in connection with a deepfaked video of Home Minister Amit Shah that went viral.

Misuse of generative AI tools in the recent past in volatile Manipur cannot be ruled out. Expressing his concern over the misuse of the AI tools in generating misinformation or disinformation in the state, Hidangmayum Shatyajit Sharma, an expert in cyber security and a research scholar at the National Institute of Technology, Imphal said,

“Although there have been no confirmed reports that generative AI was used in the recent violence in Manipur, however the fact that generative AI could be exploited in these situations is a matter of concern. Since most of the videos and photos relating to violence are circulated on various social media platforms, it is a cause for concern that generative AI, especially deepfake technology, could possibly be used to make false videos or pictures that stir up violence, spread misinformation or intensify the existing tensions between communities.”

During the past few years in India, there has been an exponential rise in the instances of deepfakes going viral on social media. This has been attributed to the absence of “comprehensive regulation in place.” What is most worrying is that there are professional deepfake creators who can be hired.

According to a BBC report, some politicians in India even got “pornographic imagery and morphing of videos and audios of their rivals” created with the sole intention to tarnish their images. Apart from creating fake images and videos of politicians and celebrities, there are enterprising software engineers who develop generative AI tools which can help a layman “create campaign materials for Indian politicians.”

Now it is so easy to create deepfakes that anyone can make doctored images or videos in a few minutes and post them on social media platforms. While talking to the BBC, Srinivas Kodali, a data and security researcher said, “And these fakes are making their way into the mainstream media.”

This has vindicated what the people of Manipur have been saying - the mainstream national media are inadvertently or intentionally spreading lies or misinformation about the violence in Manipur.