Artificial Intelligence + Internet of Things, Prototyping, System Design, Product Design,

Teamwork (2), duration: 4 weeks,

Spring 2019

AIoT Cactus is an artificial intelligence + internet of things system that helps a user, serving as a social media assistant which collaborates with smart devices, to manage and update their digital account.

SYSTEM DIAGRAM

Instead of discussing an ideal AGI that seamlessly simulates the whole of a user in the distant future, it mainly focuses on a notion of dispersal of consciousness embedded in IoT devices where different AI systems collaborate together as a collective consciousness to represent the will of user on social media.

HOW DOES IT WORK?

Hover to see the details >

The AIoT system serves as an integrated personal social media assistant. There is a pet monitor which is a smart camera that observes the surroundings and helps users to capture their important moments in their daily lives.

bubble_gif .gif

After the sensor detects a possibility, it will tell the lamp (in the right corner) to light up and the Hubble-bubble machine to make dreamlike bubbles in order to create a scene for the monitor to take a beautiful photo.

MudaneAI_codinggif.gif

Then, the photo will be uploaded to Google AutoML and wait for the evaluation. If it gets a high score, it will be passed to another AI agent,  Google Cloud Virtual Machine, to detect the content of this photo.

Screen Shot 2019-03-07 at 7.04.00 PM.png
Screen Shot 2019-03-07 at 7.04.18 PM.png
MundaneAI Proposal3_team 3.002.png

some of the evaluation results

After finishing the contextualization, the Char-RNN system will generate a caption based on the particular content. 

Jin, Miranda_Proj1_Slide6+.png
gooddemo.gif

Following is the test account on Instagram we created to hook up the whole system we proposed. All the posted photos and captions were generated in real-time.

001S.jpg
002S.jpg

Ideal collages based on the idea that

what would happen if we use stickers and filters to make our contents go viral?

004S.jpg
003S.jpg
YOLACT.png

I tested one of the yoga-cat collages with YOLACT, an algorithm does real-time instance segmentation and object recognition, which proves the idea that

we can use stickers as add-on social values to our digital existence.

Will this become a new format of hybrid reality?

Physical Installation