The Evolution of Deep Fakes: Part 1

Did you know that a UK energy firm was scammed out of $243,000, after the CEO’s voice was mimicked using #deepfake technology?

From the video of #Obama delivering a speech that he never actually did, to hundreds of conspiracy videos during the #pandemic, it is safe to say that #deepfakes have gone out of hand.

But have you ever wondered how they are actually made? Who came up with it? In the first of the two-part series on deepfakes, we decode their evolution and how they got to this point today:

Deepfakes are synthetic media created using #artificialintelligence, #machinelearning and deep #neuralnetworks. The term ‘deepfake’ refers to the use of these techniques to create realistic images, videos, and audio recordings that appear to be real but are actually completely fabricated. Let’s take a quick look at how deepfakes got to where it has today: The earliest forms of deep fakes were simple image manipulations created using photo editing software. These manipulations were often extremely obvious and easy to detect- unlike today. The term "deep fake" was coined in 2017, after a Reddit user created a tool called #FakeApp that made it easy for users to create realistic fake videos by swapping faces. One of the first deep fakes that went viral, was a video featuring the face of actress #galgadot superimposed onto a pornographic video. Since then, several other celebrities have been targeted, including Emma Watson, Natalie Portman, and #taylorswift. Things only get worse from here… In 2018, deep fake technology advanced to the point where it became increasingly difficult to distinguish between real and fake videos: remember the Obama video that went viral? It was created using a combination of machine learning #algorithms and FakeApp. It was created as a proof-of-concept to demonstrate how easy it is to create fake videos that appear to be real. The use of deep fakes is not limited to just videos. In 2019, with the introduction of "deep voice" technology, it is now even possible to mimic a person’s voice to make them say things they never did.

The use of deep fakes in political campaigns became a concern in the 2020 #usa presidential election. As the pandemic took the world by storm, thousands of deep fakes were also used to spread conspiracy theories and misinformation about the novel virus. In recent years, deep fakes have become even more sophisticated, with the ability to generate entire synthetic faces and bodies and manipulate lip movements in videos. In 2021, #openai (of #chatgpt fame) released a new deep fake model, DALL-E 2, that can generate high-quality images of virtually anything the user can describe.

As technology becomes more advanced, the threat of deep fakes to society increases, highlighting the need for better detection and prevention methods.

To combat the threat of deep fakes, researchers are working on developing algorithms that can detect them. One such algorithm is called “Fighting Fake News with AI”, which was developed by a team of researchers from the University of Maryland. This algorithm uses #machinelearning to detect deep fakes by analyzing the characteristics of the face and voice in a video.

But how are deepfakes actually made? What is the process? What are the legal and ethical issues that come with them? Stay tuned for part 2 and follow The Next Big Thing for decoding AI, pop culture and digital trends. #AI #openaidalle2 #imagemanipulation