Deepfake AI making waves: A look into its impact and potential
I have a big interest in AI and all its new and old capabilities. In continuation of exploring the ethics of big data and machine learning, I looked at the recent discoveries in Deepfake videos and audios. Needless to say, we only see this used in movies and not so much in our real lives. However, this does not mean it is not used in real life. In fact, this has been used in manipulation of news. So, in this blog I talk about ethical thoughts and considerations of using this technology.
Understanding Deepfake AI
Deepfake AI short for “deep learning based fake artificial intelligence” is AI used to create realistic but fabricated images, audios, or videos. This is a good use of deep learning (a machine learning subtopic), where artificial neural network inspired by human brain and structure, learn to perform tasks by recognizing patterns in data. Deepfake uses generative adversarial networks as well as deep neural networks to imitate real situations.
However, where do we draw the line at this amazing technology? When do we get to use it without crossing the line drawn?
The allure of Deepfake AI lies in its dual nature. It represents both incredible opportunities and scary challenges.
Concerns of Deepfake AI:
The concern is misinformation, impersonation, and privacy violation. Deepfake is often used as:
- Fake news to create fake political speeches, or news reports which could damage a politician or a group of people with same political stance.
- Impersonating someone and using them for fraud, blackmail or extortion.
- The list of uses goes on, however the vast usually violate a person’s privacy and reputation.
The biggest one to consider is that most of the time deepfake is used without the real person’s permission therefore no authenticity, no consent, and no privacy.
The line is drawn, so can we ever consider using Deepfake for good?
Deepfake AI can be best used for entertainment purposes:
Imagine you were a screenwriter or director, and you could Deepfake a screenplay just to see if a certain actor will fit your idea for the film, or even as a form of new entertainment outlet.
Or imagine if you were seeing your great-grand mother’s photo in some old tree and grass background and you can take that image and create a story your mother once told you about, preserving history and culture, taking it from a simple photo to an imaginative video and audio context.
Deepfake AI for people with disabilities:
There is a visual assistance for blind people that has been developed which is wonderful in that it can assist those who need it to be able to communicate and still stay in the know like the able people. This is the most useful case of Deepfake AI that already exists and is accessible to those who can afford it.
Examples of Deepfake AI used lightly in our daily lives:
The most common examples are from those filters in Snapchat that change your face to look different than it is, like having a beard, looking like an old lady or looking like a little child. Another example is from TikTok where you can create a video of something and alter your voice and make it sound distorted yet realistic. These are not so perfect examples of deepfake AI. None of the most accessible deepfake AI are perfect. A great deepfake example that has recently made waves is the use of an avatar. Much as this is also not perfect, however, it does consider some of our concerns.
I believe we need to investigate giving this technology a chance. If research can be conducted to enhance the best use of this technology, policies can be created in order to ensure safe usage as well as consider the ethical concerns that the majority of people have. This technology has great deal of potential, and we could be using it only in scenarios whereby it will better people’s lives and not harm them.