AI can create realistic images by itself and the results are astonishing

When we had a look at disruptive and innovative technologies that blew our minds in 2018, artificial intelligence simply couldn’t be left off the list. AI has a wide range of uses in the real world, and we (sometimes unwittingly) interact with the technology every day.

One of AI’s most interesting developments is duelling neural networks, otherwise referred to as generative adversarial networks (GANs). GANs require two neural networks that are trained on the same data set, after which one of the networks (referred to as the generator) starts to create completely new images, based on what it learnt in the data set. The second network, or discriminator, then has to decide whether the new image is real or fake.

GANs have only existed for about four years or so, but the development of the tech really is astonishing. Considering the fact that most people probably wouldn’t recognise these images as being created by a computer, this technology, refined by the American graphics processing company Nvidia, looks like it passes the Turing test with flying colours. Don’t believe us? Check out the images below.

There are a host of uses for this type of technology in the real world, and keeping in mind that GANs are also able to create realistic images of just about anything (cars, household products and even cats) stock photographers might get a little antsy about the possibilities that GANs bring. Nonetheless, it is exciting to see machine learning in action, and good to keep this in mind in an era where fake news could be getting more believable than ever before.

Ready to connect with Agora Tec?

Smartcell Logo


Agoratec Logo

                 Careers      Blog      Privacy Policy

© 2021 Agora Tec. All rights reserved.

Website by Icon Designs

Get in Touch

36 Morsim Road
Hyde Park
Johannesburg 2196


Phone: 010 006 2050