During the summer an image generator was revealed online that managed to turn even the childish of doodles of a face into one of the most realistic of images. The system used the generative adversarial network, a new type of algorithm, to create the AI faces.
AI Has Now Learned To Generate Human Faces That Are Photorealistic
Now NVIDIA has gone one better as they have come up with a system that makes use of GAN and it creates some of the most realistic looking images of people ever to have been seen.
Artificial Neural Networks Mimic The Neurons In The Human Brain
Artificial neural networks are a type of system that has been developed to mimic the activity of neurons in the human brain. In GAN, two neural networks are put up against one another. One of the networks then works as the generative algorithm and the other challenges those results by playing a role that is adversarial.
AI Images Of Faces That Are Of Unprecedented Quality
The generative network creates an image at a resolution that is lower and then the discriminator network assesses the work. With the progression of the system, the programmers then added in new layers that dealt with details of a higher resolution until GAN managed to generate images that were of unprecedented quality according to NVIDIA.
NVIDIA revealed a video of the GAN working and the faces that it created artificially were eerie and remarkable at the same time. If a person did not know that the faces had been machine generated they might have believed that they were actual living people.
The blurring of the fine line between human and faces that have been generated by machines is a topic that is much discussed in artificial intelligence, and the GAN system of NVIDIA is not the first AI system that has been able to copy something that is human.
AI Speech Now Sounds Like A Human Talking
Many artificial intelligence systems make use of deep learning techniques so that they can produce speech that sounds like a human being. The DeepMind of Google has WaveNet, and it can copy human speech almost perfectly. The Lyrebird algorithm can also copy a human voice by using only one minute of audio.
Perhaps more fascinating, or some will say disturbing, is that AI robots are now supposed to be able to understand and express human emotion. Some examples of this have been included in SoftBank’s Pepper and Hanson Robotics Sophia. We are now in the age of smarter machines and the ability of AI to be able to perform tasks that in the past only humans were capable of doing is improving. The line between machines and human beings is continuing to blur, and the biggest question now is whether that line is eventually going to disappear forever.