topos


topos is a speculative design project which addresses the possibility of ubiquitous AI systems, and anticipates the shadows they might throw on the urban landscape.


keywords
ux
research for AI
design for public space



tools

Illustrator
After Effect
Photoshop
Physical prototyping
Char-rnn



recognition:

︎presented as part the paper Right to the Post-Internet City: An Internet of Enlightened Things, at the Post-Internet Cities international conference at the Museum of Art, Architecture, and Technology (MAAT) in Lisbon, Portugal, in 2017

︎presented at the UX of AI, AAAI Spring Symposium at Standford University. And it published as paper Tree of Knowledge: Designing with Artificial Intelligence in the Urban Landscapein the symposium proceedings

︎exhibited at Ars Electronica 2017, under the umbrella of The Internet of Enlightened Things: AI in the NeighborhoodFor the exhibition, the project was reframed as AI Upkeep: Department of Parks and Recreation.



topos explores how public space can be used to demystify civic AI and ML for citizens living in an AI-embedded city. We propose a new typology of public space that combines the mechanical qualities of urban dashboards with the permeable and spatial qualities of public parks. These new “parks”, which we call topia, contain trees of knowledge that physicalize what is otherwise invisible to citizens: the decision trees, random forests, and neural nets that are involved in the public decision-making and sense-making that literally and figuratively shape an AI-embedded city.

If the topia is where civic affairs are conducted in plain sight and in real time, then tress of knowledge are both tangible user interfaces and civic symbols. Trees of knowledge provide public insight to how neighborhood data is being transformed by the various AI dimensions of the city. By externalizing civic AI systems into public space, Topos aims to create active and participatory modes of collecting data used to train these systems, as opposed to passive accumulation and infinite siloing of data.


Animation demonstrating how pruning trees of knowledge would affect the neighborhood and the city over time.



Fig. 1 CyberSyn operations room meets public park space. What if you could walk through an AI-mediated urban dashboard?


Fig. 2A Installation view of trees of knowledge in a topia.

Fig. 2B Diagram of first formal iteration for the trees of knowledge; in order to model AI systems, we created civic monument-scaled forms that manually contract and expand. The outer faces of this form serve as input layers and output layers for the neural nets that learn from city data in different ways; by unfolding and expanding the form, citizens and civic workers can respectively read and revise the hidden layers—where AI systems transform city data into intelligence.



Fig. 3 Editable trees of knowledge enable citizens to add, subtract, emphasize, or de-emphasize elements on input layers in order to create different short- or long-term outcomes on matters—ranging from self-driving car congestion to urban green space development. Here, a citizen annotates an input layer in order to change the training data for this particular AI system. Civic workers take these changes into account as they modify learning pathways in the hidden layers.
(Photo credit: Philip Van Allen.)



Fig. 4 The second formal iteration for the trees of knowledge. This visual representation is inspired by diagrams of decision trees, neural nets, and other machine learning algorithms.



Fig. 5 A speculation of the distribution of AI parks within a neighborhood. Each park contains trees of knowledge that perform specific civic functions.

Fig.6 Neighborhood mapping


Prior to Topos, I did another AI/ML project "random forest".

design questions
What does taking care or tending to AI look like in the future? What is the right amount of “tending” you need to give to an AI? What different tending methods result in different consequences?

concept
random forest materializes the machine learning algorithm by building a tangible user interface, inspired by the act of caring for plants. The AIs are shaped and grown through the act of tending and respecting. By tending AIs, such as rotating the AI based on the cycle of the day, pruning and bending the AI, you could shape and grow AIs the way you like. There are four phases of tending.


~phase one~
Tending level: 30%
Let it grow!

~phase two~
Tending level: 50%
Following the cycle of the day, rotate your AI to the side you want it to grow to face the sun! The side faces the sun would grow faster than other sides. You can grow the way you like but still retain the randomness.

~phase three~
Tending level: 60%
Pruning the part you don’t like in the growing process! Keep what you like:) It would modify the result at the real time.

~phase four~
Tending level: 90%
Bending and twining your AI the way you want it to grow. In this way, you almost have total control to grow your AI as you like.


Thanks for watching!
©Xiaoxuan Liu. All rights reserved.
Los Angeles, CA