Stories & Projects

An Experiment in Machine Learning

umbraco.MacroEngines.DynamicXml

Over the past decade, AI and machine learning have provoked debate in the global design community. But what if machine learning could be used as a tool to break down gender stereotypes and spark creativity among young children?

Text: Jens Persson / Images: Inna Zrajaeva & Prithvi Ranjan

While worries regarding the ethical application of these technologies may be valid, we are also seeing numerous useful examples in our daily lives. For example: traffic apps for your smartphone, spam filters for your inbox and even complex medical diagnostics. Machine learning is essentially about teaching computers to learn from experience, the way humans and animals do. The algorithms used in machine learning are programmed to find patterns and to keep adapting and "learning". Generally speaking, the more data you put in, the more accurate and refined the result will be.

During a two-week crash course in machine learning, IxD2 students were challenged to develop an original concept using software tools offered by machine learning technology.

"I think I've now learnt that you don't have to be a coding wiz to apply machine learning in your design process. During the course, this notion was actually quite quickly debunked. After being able to "look under the hood" of some software models where machine learning is used, I think I can now better understand when such methods can be really useful in your own designing", says Prithvi Ranjan, IxD2 student.

Challenging stereotypes through the power of AI

Inna Zrajaeva, who teamed up with Prithvi for this project, were quite surprised about where the machine learning software took their own design process.

"We were inspired by our tutors to get really creative with a range of different machine learning technologies. What we ended up doing was to feed around 400 images of various cartoonish 3D characters into this machine learning software called Runway. Then, we set it up so that it could create some new and unique characters from these images, and the output that we got was super-interesting. In the end, we got a set of blobby and really ambiguous looking creatures."

D3141E36 8F8B 456E 9534 7F52A0617633

The gender-neutral and quite fluid beings that popped out of the machine learning software inspired Prithvi and Inna to use them as empty canvases in a creative storytelling exercise for parents and kids. By involving characters that don't fit the mold of traditional gender stereotypes, Prithvi and Inna hope to counteract the figures that typically line shelves in toy stores the world over, such as overly feminine dolls and male muscular superheroes. 

The final product is a website called Tiny Tales. Here, children and parents start by randomly generating their lead characters, and then naming them. The next step is to start writing the story. After a paragraph or two, the text-based machine learning algorithms take over and start producing the next part of the story. The rest of the experience now becomes a creative back-and-forth between the kids, the parents and the machine learning software.

Screen Shot 2020 09 21 At 155952

When the adventure is concluded and the fairy tale has reached its conclusion, the programme automatically creates an entire childrens' book ready for printing, with images to go along with the story.

Adding machine learning to the designer toolkit

The course has allowed students to grasp the basics of machine learning, as well as incite discussions on the technology's potential merits and drawbacks in society. For Inna, Prithvi and their classmates, the introduction to machine learning has also added another key instrument to their design toolkit.

Runway Screen Shot

"Going forward, I think we can now use machine learning to more easily create digital design prototypes that actually work and that are more re-active and agile than a more straightforward video, for example. It can help us by giving a design idea or concept further validity at an earlier stage of the design process", concludes Inna Zrajaeva.