Fashion Bot is an imaginary Artificial Intelligence inspired by the personality of famous fashion figure in media, Anna Wintour, editor in Vogue magazine, to interact with Cher Horowitz, a fictional character who is high school fashion savvy in 1990’s movie Clueless. The project explores the creative possibility of Machine Learning(ML) to ‘unpredict’ and ‘de-create’ future fashion trends. I imagine retro-future when machine generates graphics and designs applicable to ordinary fashion product, based on one’s voice, face, mood, behavior and hand gestures to hallucinate aesthetic trend. And it further suggest characteristic of fashion AI with neurotic, edgy, harsh personality to judge one's fashion taste to teach ‘What real fashion is’ based on data. Can AI forecast design trend? if so, how can it suggests customized fashion style for multiple users? or how can AI changes the process of creating designs in creative industry? What would be the difference between human’s aesthetic standards and AI ?
In the movie, AI suggests the daily fashion style as ordinary morning routine for Cher. And, the function of the system is limited to styling. When we reimagines this system in current AI researches, how can this scenario be reimagined? In my scenario, the fashion bot(AI) analyzes Cher’s selfies uploaded in her social media account: her movement, gestures, voice and body shape, all of which quantify one’s attractiveness through the lens of high school ‘coolness.’ And, it further suggests her daily look with “overly-trained” sense of fashion.
1. Fashion AI personality through Text based ML
Various layers of data and systems of learning needs to be reinterpreted for meaningful interaction with human. I choose to use the Andrei Karpathy’s char-rnn model to train the fashion forecast and verbalize prediction of humans’ fashion aesthetic taste.
I trained with several text found from WGSN, Vogue and Elle magazines in 2017 FW.
And below is the ‘Stimulation’ text to create the output:
Fashion trend is ...
What to wear today...
2. Visualize Prediction
Other than verbal interpretation for fashion trend, I explores several AI tools: such as pix to pix, image hallucination and project muze to creates designs that can be suggested for ‘customized’ fashion products ,hallucinated by the movement of one’s hand gestures and doodles.
3. Randome factors to forecast daily fashion
Other than the collected personal data, Fashion Bot considers environmental factors to suggests styling. such as weather, temperature and geographic regions. Below is the ML that trains to predict beloved color for next week, based on corelationship between temperature in Los Angeles and most searched colors from google trend.
Brief & Context
Create a product/system concept that takes a critical perspective of design in Artificial Intelligence/Machine Learning. Your goal is to design something that on the surface seems almost plausible and sensible, but in the final analysis is useless, absurd, or really terrible.
The term “useless” can be interpreted in a range of ways, but the project must take a critical position and dig into the challenges, affordances, and potential failures of artificial intelligence and machine learning. Please take risks, don’t worry about practicality, and have a sense of criticality and humor. In particular, consider the role design plays and what it can bring to this often engineering driven domain.