Summer reflection on imagining the future of AI/ML for creative tools
by Sisi xi Yu
Adobe Machine Intelligence Design Intern, Summer 2019
Last week passed in a blur, presenting my summer project to the Adobe Design team, then flying to San Jose for the intern summit, I can’t believe it is already my last week. Before I leave, I wanted a way to remember what I learned over the summer. So here is my short, illustrated reflection documenting some of my thoughts in the past three months. This is not meant to serve as a didactic piece on how designers should approach designing for Artificial Intelligence/Machine Learning(AI/ML)nor does this refer to any specific design framework or process.
Paradigm Shift: A New Way of Looking at AI/ML Design
When I first started at Adobe, I came in thinking I would be looking at optimizing design workflows for existing flagship products. At some point in my design process, I realized that I had to wipe clean my pre-existing ideas for AI/ML functionalities before being able to think beyond the present- that there is the potential for another kind of creative relationship with AI. One of the ways that this new creative freedom is brought into Adobe’s tools is through the advancement of Generative Adversarial Networks, which can generate new audio or visual data. Currently, the user’s ability to create is dependent upon the depth of knowledge regarding the creative tool they are using. In the future, perhaps there is a chance to democratize creation using Machine Learning and GANs as a new creative medium. Instead of pixel-pushing, imagine being able to work with the machine to experiment and generate new imagery. GANs open new possibilities as a creative medium in instigating a shift in the design cultural paradigm.
Machine Learning is the New Buzzword: We Need New Vocabularies
In the age where Machine Learning became just another hashtag, it is ever more important to understand the unique properties of different Machine Learning models beyond exaggerated cliches or analogies. Perhaps it is time for new vocabularies for when working with AI beyond just “magic happens.” It involves layers of neural nets and a lot of training.
These exaggerations are partially the result of pop culture homages in sci-fi films, bold statements in media headlines meant to catch the public eye, alongside undecipherable acronyms only pertinent to those fluent in technical jargon. What I have discovered during my internship is that at a time when data privacy is at the forefront of our conversations, it is no longer sufficient to justify the lack of transparency due to the black-box model. Perhaps it is the perfect time to open up that conversation and start to communicate about what might be happening. Conversely, while the seemingly infinite possibilities of GANs are extremely exciting and intriguing, they also come with a unique set of limitations and interaction mechanisms.
Design by Experimentation: Exploring the Latent Space
Since very few guiding frameworks exist on how we should tackle these design challenges related to AI/ML capabilities, we have to experiment often. This made me especially anxious at the beginning of my internship — constantly doubting if I was exploring the right thing. A month in, I remember frantically approaching my manager on how I should tackle my internship project, he responded with“If you are really struggling, I can define the scope for you, but I wanted to give you time to discover it yourself.” I realized that to be comfortable with exploring the unknown, I had to be okay with not having the right answer on the first try and letting go of pre-existing ideas. It is a different way of designing than I was used to — much more experimental and fluid and not necessarily bound by any specific approach or framework.
If I said my design process was linear and streamlined, I would be lying. I spent the first four or five weeks exploring different options, each time realizing my scope was yet still too broad. I discovered that sometimes I would try to sit there and think things through, but my mind would draw to a blank, that is until the moment I started being scrappy and starting making mockups and testing my ideas. At some point I had a realization: if my creative process was so messy, how can the machine help myself or a potential user get to some eventual end goal in the workflow? I saw an opportunity for intelligent facilitation and sought to imagine what the feedback system between the user and machine could look like.
Value of Design: Battle of the Existential Crisis
Over the summer I was fortunate enough to have been exposed to discourse regarding the future role of design in the face of emerging tech. With the homogenization of creative tools and a time when anyone can create content, what value do designers continue to hold? I am always an optimist. I believe that the advancement of AI/ML capabilities gives designers an ever more crucial role in tackling the wicked problems we are facing.
For so long, we have been searching for a design problem to fit the algorithm. This results in deeply convoluted interactions and tech for the sake of tech. It is only natural that users feel disconnected have set off a wave of fear regarding automation. Hence, in the era of data privacy scandals, we need design more than ever in establishing trust and transparency between the user and the technology adopted. Design brings the product closer to the user, and designers frame problems in a user-centric way that can be solved. Coming from various backgrounds and expertise ranging from graphic design to political science, philosophy and film to cognitive science, computer science, and architecture, working alongside my team this summer gave me insight into the value of hybrid thinkers and designers. The diversity of background is crucial from the beginning, collaborating with the Adobe research and engineering teams to ensure that we are aware of the user’s needs, potential biases in the dataset and other potential blind spots.
Conclusion:
This summer has been transformative, not just in exposing me to the inner workings of the most prolific distributor of creative software, but also to be able to contribute to conversations surrounding the future of Creative Tech. I am an optimist, yet at the same time painfully aware of the potential negative impact of AI/ML. My team and I share a concerning number of articles on how Machine Learning is used for malicious intent, and are constantly reminded of our responsibility in considering the ethical implications of our design. I am nowhere near figuring out the answers to the questions I raised earlier, but I call for designers to start believing in our role in designing more humane interfaces in trying to design for a humane world.
Sisi was a summer intern with Adobe Design’s Machine Intelligence Design team, based in New York. She is currently pursuing a Masters of Integrated Innovation for Products and Services at Carnegie Mellon University. She has a background in Architecture and Design.
The Adobe Machine Intelligence Design team promotes a user-centered approach to machine learning and artificial intelligence in creative tools and works to establish a deep understanding of how these technologies are changing the ways we interact and create.
