Generating visual content has become a popular feature of artificial intelligence (AI). In the past, moviemakers and advertisers adopted such technologies as VFX to produce stunning special effects and shoot videos at a lower cost. AI has the potential to make their life even easier by generating visual content without any need to bring actors and cameras to a set. AI is now being actively deployed to produce all kinds of visual content including images, artwork, and videos. This article reviews five curious examples of how AI can generate visual content.
Creating a Recipe Image
A research team from the Tel-Aviv University developed AI based on neural networks that can read a recipe and further generate an image of what the final product cooked by chefs will look like. The neural networks-based algorithm developed by these scholars takes advantage of a large recipe dataset, which was uploaded to generate images of non-existent products. The visual content created by this neural network allows restaurants to understand in advance whether the finished product will look appealing to their customers even before chefs start cooking it. Nevertheless, if this technology is widely used by restaurants to post fake images on social media and design non-existent menu items, customers will quickly feel cheated and disappointed.
Photographic Portraits by NVIDIA
In December 2018, a well-known technology company NVIDIA published a paper, which detailed a new technology that could produce images of non-existent people. The researchers working for this company modified the Generative Adversarial Network (GAN) architecture to generate very realistic portraits. The modified GAN, when combined with AI, could deploy multiple machine-learning algorithms to draw life-like images at no time. NVIDIA’s new AI technology has launched a real competition between multiple machine learning algorithms developed by alternative research teams to generate fake photographic images and selects those that look best.
AI is also broadly applied to create artwork. No one gets surprised today that pieces of art produced by AI look as good as those painted by the most prominent artists in the world’s history. A Paris-based art group known as Obvious Art ‘painted’ the ‘Portrait of Edmond de Belamy’ by means of a custom-developed AI algorithm. This algorithm was previously trained on the basis of 15,000 paintings drawn by some of the top painters between the 14th and 20th centuries. In the subsequent auction run by Christie’s Auctions & Private Sales, an anonymous buyer purchased this AI-generated visual content for half a million dollars. The auction house initially estimated that the artwork produced by Obvious Art would be sold for only $10,000.
Modification of Images
AI-powered image processing service Deep Dream Generator can substantially modify an image uploaded to it and create a new one following an art style specified by the user. This service offers multiple art styles such as Deep Style, Deep Dream, and Thin Style to auto-generate images. Interestingly, several image styles can be merged by the AI algorithms powering this service to completely transform a picture. Bloggers and designers can breathe a new life into their old images using the Deep Dream Generator service.
Fake Images by ThisPersonDoesNotExist.com
ThisPersonDoesNotExist.com creates endless fake images of people that are available for free download for everyone. The algorithms that power this website were trained on the basis of large datasets of real images snapped by photographers to produce real-looking visual content. Since these fake images do not have any copyright, anybody can download and use them. However, as in the case of fake recipe images, it would be unethical to create fake social media profiles or customer reviews based on these photos.
AI is no longer a technology that lacks credibility. The ability of AI to create visual content will bring a number of benefits to companies, which include reducing costs and quick generation of realistic content. At the same time, artists and marketers need to use this technology very carefully because improper use of non-existent visual content can adversely influence their credibility and bring legal problems.
Ellie Richards is an Online Marketing Manager for Original PhD, specialising in PhD research paper help. She specialises in research, content and article writing on various topics, including Education, Marketing, and Technology.