Mary
Zhang

 
is a creative technologist based in Los Angeles. Her expanded practice involves simple robotics, olfactive interactions, creative coding, and hardware prototyping.

Her work seeks to investigate the implications of ever-evolving technology as tools, and how our labors and ecologies shift as consequence.


About
Website
Resume
Linkedin






The Scent in the Machine


MFA Thesis Project
Sept 2023 - Apr 2024

Methods  
olfactive material, electronic motors + atomizers, wax sculpture, Processing/Java, Arduino, electronic hardware 

Writing
“maybe this is all because I want to be able to smell machines.”

Showing
 Hauser & Wirth independent publishing market (2024)
 Print Pomona Art Book Fair (2024) 
 ArtCenter College of Design Grad Show (2024)

About
A speculative proposal of embedding our artificially intelligent machines with traces of the humans that inevitably fall in the loop of automation.

The ways scent 
delivers us back to our bodies, 
underscores liveness,
lingers and transports, 
allows it to carry information differently than visual signal. 

In this, scent resists reduction. How else, can the presence of their work be made known?







Click to open documentation website







Brief


How can AI-driven interactions be embedded with the contexts from which the training data is sourced?



A vast workforce of over a million workers, primarily consisting of data labelers in the global south earning $1 per hour. 










Live network atomizer, activating scents correspondingly to TCP/IP packet codes in network




Project components




Network sniffer
(link to Github)




Blended scents and motorized atomizers



Motorized wax sculptures






Click through story-game built in Processing, activating scents and sculptures with clicks and key inputs. 


An anonymized, de-identified packet sniffer (built with Carnivore) built in Processing, the live network atomizer marks the atmospherics of our networks in real time through our online activity. Our information transfer, scented.

Formulated scents based on Ghost Work and hand-made olfactive vessels connected to click-through story.  



Motorized ephemeral cast wax sculptures. 






Interactive Processing (Java) desktop interface connected to Arduino and motorized sculptures for interactive desktop story




Interactive desktop story full set up: left: motorized wax sculptures, right: electronic atomizers with custom scent blends









Network Sniffer | Carnivore




Prototype
 An exploration in materializing the invisible, the Network Sniffer is built in Processing with Java using the Carnivore library. It listens to Internet traffic on a local network and parses and visualizes network activity based on TCP/IP packets. These packet/port numbers indicate activity like Mail, SSH logins, internet http activity, chat messages, etc. 



Packet sniffer built in Processing, using Carnivore library developed by RSG, visualizing live network packets, connected to atomizers releasing associated scents. 






On AI, if the system running on data and informatics is built by us, by our activity, on our information, how can we disrupt the cybernetic apparatus and reclaim the traces of human labor in these networks and algorithms?

AI companies are raising tens to hundreds of millions in venture capital funding while the data labelers have been estimated to make an average of $1.77 per task.

















Prototyping and Testing 



Prototype
How can I build speculative sites where algorithmic labour seeps thorugh in our every day digital platforms? Repurposing familiar home and login screens of search platforms Google and Bing, I used Processing to insert input fields, buttons and evocative text prompts in these simulated screens connected to Arduinos to actuate olfactive hardware and motorized wax sculptures to release scents and signal decay.  Prompts arise when inputs are entered, provoking moments between the instantaneous where signals of turk labor could arise. 
 









Testing

 After testing with 10 users, insights were discovered highlighting gaps between the intended interactions and actual user experiences.









Insights


First-person narrative





User prompts





Providing scent contexts
  


In order to fully explain the prototype and it’s intended usage, using a first-person narrative to set a specific point of view proved most effective. This led to re-designing the screens as a linear, interactive story rather than a digetic prototype led to less confusion. 


Even with a linear story structure, users preferred direct instructions for interaction. In the re-design, direct instructions such as: “[enter your name and press return]” and “[press any key to proceed]” were placed above the input field for each screen. 


All users asked questions about how the scents released related to ghost labor. In the re-design, the narrative weaves in first-hand accounts of turkers and the contexts and locations their work is based - explicitly connecting with the scents of their context. 
















Self Publishing

Hand made printed matter shown and sold at the Print Pomona Art Book Fair (PPABF) and Hauser + Wirth’s independent publishing market.  

Two turns and three bursts flipbooks



Flipbooks (two turns, three bursts)


It smells crazy in here (machine edition)


Handbound screwpost flipbook. Flipbook frames of motorized sculptures in action, scented pages.


Riso printed booklet of compiled fragrance comment screenshots and accompanying perfume from Fragrantica comment threads. Machine edition.











Materials Research



Materiality
Scent secretes and oozes, evading discrete starts and ends, marking the constant running (working) processes, often invisible, bridging and communicating the invisible background actors.








Adornment
Maybe the lack of smell in and around a machine is the result of the sanitization technology inflicts. Like the removal of adjectives in textual processing that encode a default norm, the lack of smells signal its outputs are normal, cleaned, safe.

To smell the machine is to know what they remove, to know their processing. As relentless entities rely on the invisibilization of the labors that maintain them, smell can be the trace in the air that defies erasure, escapes sanitization.