Artificial intelligence (AI), in the digital tech world, is a game-changer, and photography and visual arts are no exception. That was then, and now today AI is so much more than just a tool for automation – it has the power to produce incredibly sophisticated, artistic (and many times unbelievable) pictures. Some AI-generated images, especially those produced by sophisticated deep neural networks (described in more detail below), can be so complex and organic-like that they are easy to mistake for real pictures. This paper discusses several categories in which AI-generated images shock the viewer, how they are made, and what technology is behind them.
Book a Demo with FancyTech and learn about AI-generated photos. Book a demo with us today.
History of AI in Image Synthesis
Create images from text descriptions or morph already existing photos using machine learning models such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs); Called GANs, which are short for generative adversarial networks, these models have been trained on extensive libraries of images and paintings so that they know what elements go into identifying an object in a scene. With the help of training, AI can in turn generate fresh images that closely resemble real photographs or detailed artworks.
Book a Demo with FancyTech and learn about AI-generated photos. Book a demo with us today.
Incredible AI-generated photos fall into two categories
1. Portraits of Hyper-Realistic Fake People
The most immediately striking example of AI in 2019 is probably how fake people are created using it for powerful graphic representations. The images of faces are produced by GANs, and the results are detailed facsimiles with skin pores, accurate lighting to simulate facial regions that cause shadows due to curvature (e.g. under-eyes), as well posing naturalistic non-extreme expressions. They often portray people so realistically that they could easily be photos of actual humans.
2. The Bleeds: Historical Figures in Modern Settings
Examples of these include the AI singularity essays and altered-history vignettes which require creativity but are decidedly anachronistic. Think about it, what if Cleopatra was dressed as a businesswoman of the modern era and Shakespeare using a laptop? Each of these images involved the AI knowing how to interpret not only what historical figure portraits or busts look like, but also existed in modern fashion and technology, creating a plausible if untrue depiction.
3. Have fun. in Worlds of Wonder and Fantasy
Artificial intelligence mixes and matches different elements in novel ways to form imaginary animals, and mythical lands with no precedent from nature. Whether that be in dragons with fur, forests lit by bioluminescent trees, or oceans twinkling like a galaxy of stars AI can push your imagination to its absolute limit.
4. Iconic Style Fine Art Created
Several AI applications enable users to give a new form and essence of reality, like DeepArt or Art Transfer by Google where you upload your photos for them to be re-done in Van Gogh style (or with patterns from any other famous artists like Picasso or Da Vinci). Given millions of said artists’ unique brushstrokes, color palettes and techniques the tools read as smudged algorithms-the algorithm recreates their works into new images that appear to have been painted by them.
5. Seamless Photomontages
The photomontage is made by blending numerous images into a single picture. Using artificial intelligence, this process can be automated – identifying how each part may be lit, and scaled and should appear in the scene to compose an image depicting surreal or impossible scenes: a whale floating above a cityscape; or perhaps a forest growing right at the center of an ocean.
Ethical Implications and Societal Impact
The eerily realistic appearance of these artist images does introduce some concerns about the power AI holds to generate them. As AI starts to branch out into the world of art, however, a whole host of ethical issues around creating images that don’t exist and what this might mean for privacy – facial recognition technology isn’t far away from painting your face onto any image it wants nowadays after all – as well as the potential implications on people’s jobs (I am going to go ahead here and assume artists).
Furthermore, the lack of distinction between AI-generated pictures and real photographs raises media integrity concerns since fake but convincing photos could have an impact on public opinion or extend to misinformation.
What will AI so mean for visual media?
As AI technology advances, the application’s capabilities to generate images will increase as well – allowing more creative and perhaps even ethically compromising uses of these tools. The applications to be unlocked are broad — from revolutionizing the film and gaming industries with CGI that does not require so much human labor, to augmenting designers to help advertisers create visually stunning campaigns rapidly and within budget.
Conclusion
The unrealistic but certainly close AI images are great indicators of how far machine learning and general artificial intelligence have come. They exhibit the power of AI to not only animalize reality but also to enable new realities as well, implying a future in which our idea of what can be real and artificial will become perhaps even more muddied. As we go forward, it will be important to modulate through these changes…innovation vs ethics and societal implications.
Book a Demo with FancyTech and learn about AI-generated photos. Book a demo with us today.