AI Generated Faces: The Future is Now
Artificial intelligence (AI) has taken over the world when it comes to automating identification of objects such as faces on social media platforms allowing users to tag people in photos quicker than ever before. Or if you live in China the implementation of an AI system called ‘Skynet’ that tracks every individual through cameras, mobile phone GPS, IP addresses, and webcam face detection during meetings. The fundamental AI technology that drives one also drives the another.
What is AI
Fundamentally AI it is just a network of statistically weighted nodes that are trained through iterative learning to find an optimized solution. The magic of AI comes in the learning process that usually happens before software deployment utilizing a large database of inputs for the training. For example, if you want AI to identify the difference between a picture of a sock and a shoe you will need a lot of reference images stored in a database to run through the ‘neural network’ or weighted statistical decision-making matrix. Each option or node in the matrix has a weighted value that is modified through each iteration until the system can guess to a high probability that the image you show it is a shoe or a sock; the same training occurs for facial recognition with features being trained and metrics between features calculated based on relative ratios. Filters a added as layers of the neural network, processing images differently for a given characteristic. If for facial recognition all the filter layers produce an accumulative answer that the image is a face, greater than any other probability then the software outputs what it is, and if dimensional ratios are assessed and confirms that it is ‘Steve’ to a high probability then this will reflect the returned answer from the system. The more filters you have and the larger the learning dataset is before selling the software using it the better accuracy the solution will have. In future it is likely that these software metrics will become a major selling point for apps that use AI under the hood, as greater accuracy through calculation and iteration should lead to more pragmatic simulations.
AI Generated Faces and Potential Uses
How are people using AI Video, well one use is using a ‘driver’ video that maps a user’s facial features onto a static image in real time and renders a new version of the image from it. What does that mean, well you could take a picture of Abe Lincoln and animated it with your face driving the expressions using AI in real-time. Teaching history just became much more enjoyable especially for distance learners. Alternatively you could use it to make relatives that have passed away to appear to be living; this may be useful for elderly people with memory challenges who are missing their partners. People from yester-year could be used for marketing campaigns through the production of new content. War memorials may have faces of the honored animated to help keep memories alive and help never forget the past.
You could also make interesting content videos for social media or YouTube based on history or personal memories captured in photos; the sky is the limit to the possibilities. Personal social media from decades past photos may be reinvigorated through a user’s animation. Friends may also create animations and real-time content as their audience to entertain them in a new dimension of social interaction.
Conversely this technology does come at a price as with any other. What if you take a picture of your spouse and use it to unlock their phone for nefarious reasons using a second screen. People could bully others, disrespect the dead and help fabricate fake news to spread social unrest.
Hopefully however good will prevail and the world becomes a little bit more interesting to live in with stronger social and family ties created through the use of AI generated faces with imitation as the best form of flattery. Loved ones will be seen again in a personalized and loving way and history will come alive in a multitude of settings.
