As the world continues to be fascinated by human-like AI, the conversation around deepfakes and their impact on the world has begun resurfacing.
But what exactly is a deepfake?
A deepfake is an AI programmed to replace one person’s likeness with another when in recorded video. The AI does this through what is called deep learning. This is when the AI teaches itself how to solve problems based on a very large set of data, and thus, learn how to swap faces in video to make realistic-looking fake media.
A deepfake AI will use neural networks to capture a human face and then perform a face-swapping with another when commanded to do so. So, a person wanting to create a deepfake will need a target video to use as the basis of the deepfake and then a collection of clips of the person whose face you want to insert into your fake video.
What deepfake is not
Deepfake is not the simple act of photo-shopping one’s face over another nor is it a video that has been reworked through video editing.
Instead, to create a deepfake, you need specialized apps and algorithms and a series of clips so that you create an entirely new video.
So, deepfakes are a lot more complex than simple video editing, which is what makes them so potent.
Applications of deepfakes
As we have seen, the fact that deepfakes can produce altered videos that look very authentic means that they can be used for bad purposes. A major issue is the use of deepfakes in pornography, where faces of famous female celebrities were swapped onto a porn stars’ faces.
However, positive applications of deepfakes include:
- Allowing people who have lost their loved ones to interact with their avatar.
- It could also be used in filmmaking thus saving money (it’s cheaper than VFX), time and labour.
- AI generated deepfake people can be used to create engaging customer service and save time and money.
What is the future of deepfake?
Currently, deepfake still isn’t mainstream, but with the fast advancement in AI, it could very well be mainstream within the next five years.