Tales of the Talking Dead

5 min readOct 18, 2022

Deepfakes are synthetic media in which a person in an existing image or video is replaced with someone else’s likeness. The term “deepfakes” is a combination of “deep learning” and “fake”. Deepfakes use potent machine learning and artificial intelligence techniques to edit or synthesize visual and audio information that can more readily fool, even though the act of producing fake content is not new. The primary machine learning techniques for producing deep fakes are based on deep learning and require training generative neural network topologies like autoencoders or generative adversarial networks (GANs).

What is AI

Artificial intelligence(AI) makes use of computers and machines to mimic the problem-solving and decision-making abilities of the human mind. It allows machines to model, and even improve upon, the capabilities of the human mind. From the development of self-driving cars to the proliferation of smart assistants like Siri and Alexa, AI is a growing part of everyday life.

How Deepfakes use AI

An autoencoder is a form of neural network that is used in deep fakes. These are made up of an encoder, which shrinks an image to a latent space with fewer dimensions, and a decoder, which builds the image back up from the latent representation. Deepfakes make use of this architecture by encoding a person into the latent space using a universal encoder. Important details about their face characteristics and body posture are contained in the latent representation. A model that has been trained particularly for the target can subsequently be used to decode this. In other words, the latent space’s representation of the original video’s facial and body traits will be overlaid with the target’s specific information.

This architecture can be improved by adding a generative adversarial network to the decoder. A GAN develops an adversarial relationship between a generator — in this example, the decoder — and a discriminator. The discriminator tries to detect whether or not the image is generated, while the generator generates new images using the latent representation of the original material. As a result, the generator produces exceedingly accurate representations of reality because any flaws would be detected by the discriminator. In a zero sum game, both algorithms continuously get better. As a result, deepfakes are challenging to stop because they are continually changing; whenever a flaw is found, it can be fixed.

Applications of Deepfakes

Un’emozione per sempre 2.0, a video artwork by multidisciplinary artist Joseph Ayerle, was released in March 2018. (English title: The Italian Game). A synthetic Ornella Muti from the 1980s movies was produced by the artist using Deepfake technology, and she was able to travel in time from 1978 to 2018. This piece of art was mentioned by the Massachusetts Institute of Technology in the study “Collective Wisdom.” The artist investigated generational reflections and issues surrounding the function of provocation in the field of art by using Ornella Muti’s time travel. Kendall Jenner scenes were utilised by Ayerle for the technical realization. The AI-generated visage of Ornella Muti was used to replace Jenner’s face in the program.

Deepfakes can also be used to enhance content. Intellectual giants of the past can be incorporated within videos to “teach” the ideas they wrote. This would share their knowledge more interactively

The dark side of Deepfakes

In 2017, Deepfake pornography was widely distributed online, especially on Reddit. In many online deepfakes as of 2019, women celebrities’ likenesses are usually employed in pornography without their consent. 96% of all deepfakes online, according to a survey released in October 2019 by Dutch cybersecurity business Deeptrace, were pornographic. A Daisy Ridley deepfake originally attracted attention in 2018, among other things. British and American actors made up the majority of the deepfake subjects on the internet as of October 2019.The majority of the South Korean subjects, who make up around a quarter of the topics, are K-pop stars.

DeepNude, a free download for Windows and Linux, was made available in June 2019. Its goal was to remove clothing from photographs using neural networks, more specifically generative adversarial networks.

Deepfakes can be used to produce extortion documents that falsely accuse a victim. Deepfakes could be used to blackmail political officials or anybody with access to classified material for espionage or influence campaigns, according to a report by the American Congressional Research Service.

Alternately, since the fakes cannot be consistently separated from authentic materials, actual blackmail victims can now assert that the real artifacts are fakes, giving them a convincing explanation for their actions. The result is that traditional blackmail tactics lose their validity, which erodes victims’ loyalty to the blackmailers and undermines their power. This tendency, which “devalues” actual blackmail and makes it worthless, is known as “blackmail inflation.” This blackmail may be produced using simple GPU hardware and a short piece of software.

The Zindi Hackathon

Hackathons are a fantastic way to explore, collaborate and work toward solving a problem within a defined period of time. Most hackathons explore software-related solutions in technical events whereby people, especially computer programmers and software developers, meet to engage in collaborative coding with an aim of creating usable software.

Jomo Kenyatta University of Agriculture and Technology, in partnership with Zindi and the mighty Jenga school have organized a Hackathon. The hackathon, which will be held from 21st to 25th November 2022, will be unique in that the participants will be mostly JKUAT Bachelor of Computer Science final-year students. The hackathon is aimed at fostering solutions addressing the Sustainable Development Goals by leveraging the use of Artificial Intelligence.

An interesting idea for the hackathon would be to utilise Deepfakes to create videos involving accomplished academics of the past. This would greatly improve the quality of educational content served to students.

Parting Shot For all those AI enthusiasts who also want to be a part of this competition, join the hackathon here 👇👇

https://zindi.africa/competitions