“Design is the intermediary between information and understanding.”
Technology experts claim that the acceptance and growth of any new technology is a slow progression. It’s a slow income time, placid and shallow at first, slowly growing until the degree of acceptance surges, suddenly becoming a wave that floods everything in its path. The internet grew out of military communication concerns, and blossomed from small utilitarian protocols into the complex living organism that now dominates culture and commerce. Today, we are walking in the shallows of artificial intelligence technologies, and small waves are gently lapping at our ankles. However, those who understand AI know that the tsunami is on its way. At the very least, the changes resulting from AI will be fully enveloping for designers and creators. If we don’t understand or learn how to manage the AI that is currently being incorporated into communications and creative tools, it could undermine humanity’s role in controlling in creativity as we know it.
Artificial intelligence is more than just complex mathematics— it is a process of securing information, turning it into optimized data, and using algorithms to find the best prediction. That best solution is then used to effect some desired result, guiding the car to a desired destination, and navigating the next turn safely. Choosing the best background for an image may involve turning a 2D image into a realistic 3D rendering and, in the process, turning a mere snapshot into a false reality. These are the obvious, open, and sometimes notorious uses for AI; it is the hidden or more subtle uses of the technology that create its seductive powers. The AI chip in your mobile device guides you on your walk, chooses the best images to post, and tells marketers what ads to show you. Social media apps determine who, or what, will be interesting based upon your purchases, browsing histories, and past interactions. All these functions ride on top of artificial intelligence engines. But AI is not limited to social media technologies— they are just the low hanging fruit for AI creators. The real magic is evident in computationally complex apps, such as applications commonly used by designers.
Design programs are starting to use AI extensively. They’re most visible in magical completion of the missing parts of an image, smart deletion of unwanted backgrounds, auto color, auto exposure and image sharpening tools we use. Designers and photographers have become extremely reliant on these aides, and in response, the way we design and take photographs has changed. It is a self-reinforcing cycle. Adobe, Google, Apple, Facebook, Twitter, and Instagram, among others, rely on AI to make their products easier to use. This simplicity of use comes with a complex trade-off. The data and access to information you give to cloud-based app providers is much deeper than most people understand. This access may even extend to data stored on your hard drive, on the cloud, and even on your mobile device. AI is always searching for data and the creators of AI-enhanced technologies are very creative in the way they consume your data. Why? More data makes the reach with consumers deeper; it may make the product better, more seductive and, undoubtedly, it makes the provider more valuable. However, the use of AI has an echo chamber effect— not only does it affect outcomes, it also shapes decisions regarding inputs. What happens when effect shapes the design brief? I believe AI will eventually become a more commonplace tool for making business decisions. When this happens, AI will begin to influence and shape the creative work product. It will start slowly, but at some point, business managers will likely rely on AI to make even more decisions. But as AI gets better, does that make it any more trustworthy?
Business owners want to ensure the process of creating, designing, and marketing products and services is efficient and, above all, results in measurable profits. Shiny new concepts and tools are always welcome, and few are as new and shiny as artificial intelligence. The process of classic design thinking relies on securing information about customers and understanding their needs, followed by ideation, prototyping, testing, and reiteration. Current AI technology is a perfect fit for the first and second part of this process. As the technology improves, the remaining processes will easily be incorporated into AI design processes. The problem with AI in this context is that it must rely on what it learns, and it only learns from the information that is given. Typically, the humans providing this information are not artists or designers. They are low-paid assembly line knowledge workers who make decisions based on a lowest cost basis. Humans then construct the algorithms that tell the computer what data to use. Designers have little to no influence during this process; they just have to live with the results.
Fields that rely on AI tools include transportation, industrial operations, banking, communications, manufacturing, and medicine. But AI is affecting every industry, and design software is not immune to this trend. Adobe states that their new analytics software, Adobe Sensi, is a set of tools will help designers “optimize and scale user experiences” with “real-time intelligence” and help marketers predict customer behavior based on “attributes, differences, and conversion factors.” Stated simply, what Adobe and similar companies are promising is that AI will create shortcuts, like an easy path from design to market success. But designers should remain mindful that creativity, good artwork, and good design are inherently human pursuits. Will these tools create a valid shortcut in the creative process, or hinder its natural progression?
So when, and how, should we use AI to enhance creativity? While it can be a helpful addition to a design toolkit, designers must understand how it can influence creative processes as it becomes prevalent. AI creates the promise of easy answers, or at least, a faster way to get usable solutions. For a product manager or business owner, any tool that makes it easier to understand customers and their desires is a good thing. The problem with AI in design is that, because usable data is difficult and very expensive, tool creators will be tempted to use the same data sets repeatedly. This is dangerous, as the overuse of certain data will inevitably create bias in the algorithms guiding AI. While the information within an AI solution is what creates its magic, it’s also a significant part of its danger. At the very least, resulting designs will soon lose their distinctiveness.
An artist / designer who chooses to work with AI must remain mindful of the fact that it is not one single technology. Facial recognition, gaming, and many creative uses of AI use Generative Adversarial Network (GAN) networks, a type of machine learning that is also used to create deep fakes. To my knowledge, the issue of licensing a person’s image and subjecting the photos to GANs manipulation has not yet been addressed. If a designer is working with GANs technology, the model release should at least identify that the photos may be computer manipulated.
Deep learning is a subset of AI machine learning that incorporates additional neural networks. This technology is usually associated with automation and “teaches” an AI application to make better decisions for performing analytical and physical tasks without human involvement. This technology could also train a design application to use tools in the designer’s own style or manner of working. If these applications are cloud-based and served to your workstation on demand, review the settings to ensure that you are comfortable with the default level of sharing. You may wish to avoid granting excessive access to the information you’ve created while working with the application.
Convolutional Neural Networks (CNNs) and their cousin, Recurrent Neural Networks (RNNs) are another subset of machine learning. CNN technology is commonly used for image classification (identifying an object in a picture) or feature recognition (identifying patterns and voices), while RNN is associated with speech recognition tasks. Designers who create work that incorporates image and/or voice recognition, such as UX and UXI, should be aware these technologies require specific methods of input which may affect the final UX experience. Designers will also find that computer processing power will impact the possible implementation of their designs. Designers who create UXI for reinforcement learning systems, such as teaching machines that manage large data sets, must ensure they understand the limitations of the technology. In this case, it’s important to ask questions about the type of user, as well as how the design will be used. Find out if the computing devices and human interfaces have input or graphics display limitations. You can always ensure a better solution by learning about the intended uses for a design that interfaces with AI.
As AI tools become more sophisticated, they will certainly go on to influence choices in prototyping, fonts, color ways, image styles, and design element placement. In sum, design decisions will become based upon data selected by third parties who may or may not be designers. Choices by AI engines will not evince the creativity, imagination, and exploration that all good designers exhibit. All of which leads to the following questions: who is selecting the data? What is their design experience, and what are the criteria for selection or exclusion of information? If there are weaknesses in these choices, they contribute to the bias inherent to the algorithms. If this consideration goes unchecked, AI will corrupt creativity and design. The danger of such shortcuts will result in questionable choices being “baked into” a project, and will be part of many projects when an AI engine is used repeatedly. The results will feel devoid of the search and discovery that form the uniquely human elements of creativity. The wise designer will come to understand that while AI is a design aid, it is not a solution to design problems.