Everything has pros and cons and technology is no exception. Though the internet has ushered hyperconnectivity and hyper information, shrinking the world with its impact, nothing seems real anymore. Does it sound weird? You can morph images and doctor videos easily. Celebrities who have never indulged in pornography, find their videos over the web. You can see videos of politicians showing and doing things they actually did not, go viral, creating fake news.
The beginning of Deepfake Technology
These “Deepfakes” is attributed to deep learning which is a specialized part of machine learning. This is the latest AI (artificial intelligence) branch that is based on learning data representations like speech, audio and image recognition. It is heartening to think that these vital instruments for building trust and relationship can easily toy for unethical purposes. It can destroy any reputation.
Deepfake for Image
Researchers at Stanford in US and MAX Planck Institute and the University of Erlangen-Nuremberg in Germany had collaborated to build the first software-based on Deep learning called Face2Face for facial recognition. It captures the facial expressions of the person on the webcam and morph them on the face of a person on the YouTube video. The target person in the video can be made to say things that he or she did not actually speak, so it is actually putting your words into any mouth. This is copyright infringement and sometimes even computers fail to detect good fakes.
Read More: Blockchain Technology – A Major Player of the Tech Space
Deep learning Tools for Video
Deepfake videos were first created by Fake app. DeepFaceLab hosted in GitHub is another ML program to enhance facial re-enactment and facial replacement in videos. These tools construct a high detail, 3D face model from an image which is transferred on the target person on each frame of a video.
Deep learning tools is open-source software. This means it is readily available to anyone. You need a good computer and a proper graphic card to use, collaborate and enhance them.
Deep Learning for Speech
Machine learning has led to a new series of speech synthesis technology that lets anyone use someone else’s voice easily. You need to simply copy the voice using the voice imitation algorithm. You can then make the target speak anything you want, even a hate speech which he/she had never uttered.
Lyrebird has launched its app for iPhone users where they can easily copy a voice and play with their new voice avatar. Due to ethical reasons, Adobe’s VoCo, the speech synthesis technology was not launched in the market.
Read More: Why must children listen to nursery rhymes – does it help?
The custom sex robot is developed with this latest AI technology. The client provides an image which is used to create a 3D face model and then a silicone or TPE body is molded accordingly. Most customers create their silicone wives after their favorite star.
It is disturbing to see how deep learning has turned into a weapon for mass disinformation. Facebook has collaborated with developers around the world to launch fake checking software and browser-based authenticity applications.
However, we often hear of fake videos going viral and creating panic in the society. It is time, we professionals join hands to spread awareness about fake videos and morphed images.
Read More: Why You Should Design A Math game with your kids?