AI Creates Fake Humans That Never Existed In The Real World

Nothing is scarier than the headline ‘AI Creates Fake Humans’. I am afraid that this headline is no longer a fantasy. We had earlier informed you about how AI can predict your death, read your mind or even worse — inherit our racial biases.

However, a new and evolved artificial intelligence is on the horizon whose potential for doing something terrible is enormous. This AI creates fake humans inside a computer. The worst part is that these fake models are almost indistinguishable from the real ones.

AI Creates Fake Humans

The AI is developed by a startup called Datagrid, located at Kyoto University in Japan. The new AI can create an endless number of fake humans that never existed in real life. Thus leaving us to question — ‘what is the truth anymore?’

The super-intelligent AI was created using a Generative Adversarial Network (GAN). It consists of two AIs which try to beat each other. In this particular case, one AI was generating images of fake humans while the other one was tasked with finding out the fake ones.

GAN Definition

A generative adversarial network (GAN) is a form of unsupervised machine learning system. In this system, two artificial neural networks compete against each other in a zero-sum game framework. This technique can generate photographs which look real to the human eye. (Wikipedia)

Initially, the AI generating the images was unsuccessful; however, with each cycle, it became smarter. Eventually, the AI models became so real that even another AI designed to spot fake images couldn’t tell the difference.

AI Is and Not Dumb

We recently told you how an AI could be fooled using a patch of colored cloth, the same can be applied to Datagrid’s AI. It can be seen in the video that models are standing in front of a white background, making it easy for the AI to avoid any glitches while generating images.

AI may be able to calculate the probability faster than humans, but it still sucks at recognizing patterns. Elon Musk’s famous driver’s assistant AI has been unable to avoid several fatal crashes.

On the other hand, AI is making strides in creative fields including writing stories, composing music, and even generating original puns. Although most of the AI-written pieces are still at the elementary level, they are learning fast. But there is also OpenAI, a Dota 2 competing bot who is able to defeat professional esports players with ease.

There is also the important issue of making sure that AI does not inherit our own racial, gender or ethical biases. However, with the lightning-fast development in the field of AI, the already significant problem of fake news is bound to become more of a nuisance.