Artificial intelligence has become a hot topic and more and more people are learning about it and using it in different fields. For example: painting, writing, clinical, biological, etc. The combination of artificial intelligence and art has led to a diversification of art forms and even a dramatic change in new media art. For example, the first AI-generated artwork, Portrait of Edmond de Bellamy, sold for $432,000 in New York in 2018. Here is a link to this artwork:Edmond de Belamy – Wikipedia
‘Edmond de Belamy’, a portrait generated by artificial intelligence, was sold by a collective called Obvious for over US$400,000 in 2018. (Wikipedia pic)
New media art is different from traditional art. It uses some technology to bring the paintings closer to life, and can even create the desired work according to human instructions, from 2D to 3D to imitating the works of art masters, which can be very intelligent. But AI painting also has its drawbacks. This blog will use the existing AI Chat GPT, Stable Diffusion, Midjourney and the soon to be landed AIGC in China as interpretations to talk about what are the advantages and drawbacks of AI in painting applications.
Artificial Intelligence Techniques
The first key technology for AI painting is Generative Adversarial Networks (GANs). GANs consist of two neural networks, a generator network and a discriminator network. The generator network is responsible for generating the image, while the discriminator network is responsible for determining whether the image is real. These two networks play off each other to gradually improve the image quality of the generator and eventually produce a high quality image. And then we need these two neural networks to help each other and finally add some generative conditions, such as labels or descriptions of text, and then this technique talks about filtering out the style you want and automatically generating an image based on that style or processing it into a specific image based on specific needs. Style apologetics allows the style of one image to be used for another, and then the AI uses algorithms to generate a final image that meets the criteria or creates a new work of art. These are a great help in painting, which we will talk about next. So what are the applications of AI that are so technically advanced?
AI painting is used in many areas such as design, as we will see below, or in the creation of painting art, games, AI painting is used to generate sculpture, photography and film effects, as I mentioned above, 2D and 3D effects.
Artificial intelligence and painting
Traditional painting requires the painter to think about the work or have certain feelings about what he sees in order to paint it, and it takes a lot of time to paint a picture, usually from the beginning to the end of the design, which can take as little as a few hours or as much as a few days or even months to produce a satisfactory work. Even the emergence of artificial intelligence has not only promoted the development of technology, but also the development of painting. From teaching methods to aesthetic ability to continuous reform and innovation, for example, more and more people have begun to accept the idea of virtual clothes, virtual images, and even meta-universe, more and more virtual products are being collected. This is the good change that AI has brought to painting, but what about the shortcomings? Of course there are drawbacks. First of all, artificial intelligence only works according to instructions and generates the instructions given to the computer by humans, and does not have its own ideas and aesthetics. Artificial intelligence can improve, but it has no aesthetics and cannot even adjust the details of a painting.
In terms of business models, a work of art has to have its own value, and the value that comes from creating digital technology is adaptable and agile, even unique, which is what AI brings to the table. But artificial intelligence is not fully intelligent. As I said before, AI digital technology does not have its own aesthetics and ideas, but only imitations filtered through large amounts of digital information. So when it comes to art and painting, there is still a designer needed behind the scenes to give instructions, or even to make adjustments and refinements after the AI has created the work. When it comes to commerce, there is the issue of copyright. Firstly, the digital artwork produced by AI, whether it is paintings or virtual objects such as clothing, avatars, etc., are currently copyright free, except for those products that are purchased at a significant cost. The reason for this is that the products provided by AI are in bulk, and AI is simply going through them and making some changes to create new products.
In summary, AI is only a part of art and not art itself. AI can help one understand art better and even broaden the style of art。
ChatGPT is a natural language processing tool powered by AI technology that allows you to have human-like conversations and much more with the chatbot. The upgraded Chat GPT-4 Open AI is also responsible for creating the popular DALL-E-2 AI art generator and the Whisper automated speech recognition system. Chatgpt analyses the interaction based on the context of the chat and completes the writing of various requirements, in drawing it can do copywriting and even scripting for film and television for new media. Chat GPT can constantly integrate human ideas in its communication with people, constantly concretize the objects that consumers need, and even understand complex and abstract needs, such as conveying emotions, or an object that has never been invented but needs to be represented in pictures. This enables the most difficultly to understanding and expressing artistic needs to be effectively communicated between people and digital media, artificial intelligence, further facilitating the better representation of digital media in art.
Stable Diffusion is a deep learning, text-to-image model released in 2022. It is primarily used to generate detailed images conditioned on text descriptions, though it can also be applied to other tasks such as inpainting, outpainting, and generating image-to-image translations guided by a text prompt. The more specific the description, the closer the image will be to your needs. For example, a 19-22 year old girl sitting in front of a computer. The keywords are: 19-22 year old girl, in front of a computer, sitting, and the system will search for the image you need among hundreds of images based on the keywords. But it only serves as a filter and does not include any artificial intelligence.
Stable diffusion online
Midjourney is an artificial intelligence program and service created and hosted by a San Francisco-based independent research lab Midjourney, Inc. Midjourney generates images from natural language descriptions, called “prompts”.
AIGC is an upcoming Chinese AI art creation application. It includes tools such as Chatgpt, Midjourney, Stable Diffusion, as I said above, Stable Diffusion can generate text images, also images can generate images, or improve the accuracy of images, its important plug-in called Controlnet can improve the effect of raw images, realise line drawings to full-colour images, improve the detail aspect. ChatGPT can interact with consumers according to their context, but it does not complete the work of writing various types of scenarios.Midjourney is based on the text you want to get, through the AI to generate the corresponding picture, the exquisite degree of draft, but the operation is generally controllable. So the combination of the three AI platforms to get the AIGC. its process is expected to be, human proposed needs analysis, then artificial intelligence to find thousands of references, and then manual design, and finally to the designer for details to adjust, to complete the final work.
For example, in the case of the gift ICON design, AI painting has greatly improved the accuracy of the image and has increased efficiency by 40%-50% compared to the usual process. The first step is demand analysis and refining keywords. The second part AI generates images, gives keywords, text generates images and mass produces a large number of design drawings.The third part is manual fine-tuning, where the designer makes local adjustments to the AI drawing.
Disadvantages of Artificial Intelligence in Painting
Firstly, the various concept and character drawings generated by AI cannot be used directly, and the resulting works are extracted and identified from various imagery on the Internet and then processed.
Secondly, the images generated by AI come from a patchwork of images with no ideas of their own, or even a unified plan and design, and the works generated from a single element are very different. Another point that needs to be made is that AI relies on the material and is therefore limited by it, and therefore cannot create more unique works, as the algorithm is limited by the data on which it is trained.
Thirdly, there is the technical aspect. Although AI is now at a very advanced stage, AI algorithms are subject to technical errors, such as glitches or mistakes, and it is likely that the tutor will end up with a result that is not what we want. As with my picture above, the image given by the AI is monstrous. It does not fit the human aesthetic.
The fourth point is that it also lacks emotional resonance. Humans are a group that thinks and feels, but AI is not. AI is based on algorithms, a combination of 1s and 0s. Humans’ own images resonate with viewers and consumers, whether they are happy, upset, sad, angry, etc. But AI is likely to lack the emotional depth and connection that painting can create.
So is AI painting really intelligent? Probably not. Although AI can create some realistic and impressive paintings through algorithms, no matter how we use AI and what we hope to achieve, it is only acting as a tool and not really painting. AI for painting is not yet fully autonomous and usually requires human input to guide the algorithm’s decisions and make creative choices. In some ways, AI for painting can be seen as a tool or collaborator for the human artist, rather than a substitute for human creativity.
Stable diffusion online. Stable Diffusion Online. (n.d.). Retrieved April 11, 2023, from https://stablediffusionweb.com/#demo
Holmström, J. (2022). From AI to Digital Transformation: The ai readiness framework. Business Horizons, 65(3), 329–339. https://doi.org/10.1016/j.bushor.2021.03.006
Sjödin, D., Parida, V., Palmié, M., & Wincent, J. (2021). How AI Capabilities Enable Business Model Innovation: Scaling AI through co-evolutionary processes and feedback loops. Journal of Business Research, 134, 574–587. https://doi.org/10.1016/j.jbusres.2021.05.009
Zhang, H., Bai, X., & Ma, Z. (2022). Consumer reactions to Ai Design: Exploring Consumer Willingness to pay for AI‐designed products. Psychology & Marketing, 39(11), 2171–2183. https://doi.org/10.1002/mar.21721
Wikimedia Foundation. (2023, April 11). Artificial Intelligence. Wikipedia. Retrieved April 11, 2023, from https://en.wikipedia.org/wiki/Artificial_intelligence
Tian, H. (2022). Application and analysis of Artificial Intelligence Graphic Element algorithm in Digital Media Art Design. Mobile Information Systems, 2022, 1–11. https://doi.org/10.1155/2022/6946616
White, C., & Matulionyte, R. (2019). Artificial Intelligence Painting the bigger picture for copyright ownership. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3498673
Lin, Y. (2020). Research on application and breakthrough of Artificial Intelligence in art design in the new era. Journal of Physics: Conference Series, 1648(3), 032187. https://doi.org/10.1088/1742-6596/1648/3/032187
Liu, X. (2020). Artistic reflection on artificial intelligence digital painting. Journal of Physics: Conference Series, 1648(3), 032125. https://doi.org/10.1088/1742-6596/1648/3/032125
Guo, C., Lu, Y., Dou, Y., & Wang, F.-Y. (2023). Can chatgpt boost artistic creation: The need of Imaginative Intelligence for parallel art. IEEE/CAA Journal of Automatica Sinica, 10(4), 835–838. https://doi.org/10.1109/jas.2023.123555