seams AI
AI and machine learning systems have seeped into every aspect of our lives. However, we have a twisted perception often treating it as a chimera being. There's a loss of control and trust rendering experiences that feel like wandering in a black box or walking on a minefield feed. Despite all of this, people currently tend to train everyday AI whether playing a song, clicking, hiding scrolling a feed. Our existing approach to AI-UX is not designed for feedback. In short, we need a new language to talk to AI and as designers new ways of prototyping with the statistical nature of machine intelligence. I explore how we can change this relationship with everyday AI by leveraging our existing ways of training.
How can we make our AI more naturally controllable, explorable, and adaptable for everyone?
Seams is a product that sits on top of all your AI-based applications. It guides you through the process of training your very own personal AI either individually or collaboratively. It provides an essential tuning toolkit (Curation- influence from people, algorithmic range, keywords for composition, ranking or grouping suggestions, collaborative drawing, negotiating machine errors).
A vital part of the training is representative and unbiased data. Seams compass helps you to explore beyond your bubble, adding new cultural lenses, and letting you save explorations along the way. It proposes a new interaction paradigm for navigating a feed.
AI can never run without data. Seams has a contextual data control that can be accessed anytime. These controls help users to negotiate utility vs. privacy throughout the training process.
Seams is the first product developed out of Soft machine: a pattern language for Human-AI interaction developed during the thesis.
During the project, I collaborated with Machine Teaching Group, Microsoft Research, and various industry-based AI experts working on personalization and applied Machine Learning research. This was further augmented with design sprint, co-creation with end users and tangible/digital sketches to conceptualize various unintelligible concepts. If you are excited to know about the future of everyday AI- Human interaction and ways to design and prototype with Machine learning, Join in the zoom or email me: sahooshib1@gmail.com.
Shibashankar Sahoo
In collaboration with:
