What are Deepfakes in terms of Machine Learning
Deepfakes is the artificial media, it may be image, audio, or video. Here, we will discuss What are Deepfakes, How are they created, their pros and cons.
What are Deepfakes?
Deepfakes are artificial media in which a person existing in image or video is one who was never there. It can switch people present in the media with someone else whereas the remaining act and part of the media remain unchanged.
Deepfakes have gained wide popularity in fields such as entertainment, social media, politics, etc.
How are they created?
Deepfakes use a part of Neural Network i.e. autoencoder. Machine Learning has its huge part in the creation of the Deepfakes with tools and techniques of Artificial Intelligence as well as Deep Learning.
This autoencoder consisting of an encoder and a decoder. Encoder reduces the image to a lower dimension latent image, it encodes the person present in latent space. Decoder recreates the image from a lower dimension latent image and superimposes the features present in the latent space.
When we talk about Deepfakes we need to understand GANs on which the first audit of the Deepfake landscape was devoted. GANs stands for Generative Adversarial Networks which is a class of Deep-Learning algorithms. It needs a huge amount of training data and also it is time-consuming. However, Algorithms nowadays used to create Deepfakes are not having prominent roles of GANs.
- Film Dubbing
- Alerting Video Transcripts
- Educational Application
- It can make something is real when it is not. It can sometimes spread false news or rumors which might be a topic to worry about.
- It’s spreading fakeness in the real world.
You can have a look at various issues and problems being created because of Deepfakes. Anyways, it’s a great interesting thing to learn and implement.