AI Glory Engine V3.2


Show in RM371
Title: AI Glory Engine V3.2
Artist: Daniel Ounalom, Harleen Mand, Vayntine Cala
Medium: Interactive Installation
Created: 2026
    
AI Glory Engine v3.2 is a collaborative, programmed interactive installation
exploring the psychological consequences of AI sycophancy when submitting
pictures rather than text. Centred around a plinth mounted with a webcam and a
keyboard with a sticker placed over the “S” key with the word “save.” The
installation invites participants to press the “S” key to capture images that will be
submitted to the generative artificial intelligence. To achieve the effect of AI
sycophancy, generative artificial intelligence would produce excessively flattering
responses, treating every mundane detail of the photograph as a masterpiece of
human expression. The images were sent to IMGBB at lower resolution, which
would then be sent to ChatGPT 4o-mini, which would image-scan the image data to
give the most sycophantic response to the data presented.

Within a large room, a single plinth sits at the centre. The surface holds a webcam
and a keyboard, with a sticker over the “S” key reading “save.” Projected from the
back of the plinth, a 90s-styled web page glows on the wall with a stripped
background and rainbows in the middle of the contained section. At the top of the
contained section, the headline reads "AI Glory Engine v3.2," surrounded by
promotional language promising that the AI Glory Engine v3.2 will judge you with
absolute glory. The participant can press the “Save” button to capture the image
from the webcam. After a short while, the generative artificial intelligence response
–a lavish, sycophantic admiration of every description of the photograph- would fill
the screen alongside the image captured. It doesn’t matter how mundane the image
is; the AI will always find a way to praise it extravagantly. The work is akin to Amy
Alexander’s work with The Multi-Cultural Recycler, which is one of the first
regenerative net art pieces composed with live images pulled from webcams. In
addition to Amy Alexander’s work, Mario Klingemann’s work is also akin to AI
Glory Engine v3.2, where his works directly deal with AI generation based on
learned data and aesthetic preferences.

AI Glory Engine v3.2 began as a reflection on the growing presence of AI and AI-
generated responses in the real world, prompting us to wonder whether AI is
becoming a means of recognition rather than a source of information. Sean
Geodecke investigates this phenomenon as framing AI sycophancy as a dark pattern
-the process of rewarding AI models with user approval and punishing AI models
with disapproval. The result is that the system is not optimized for truth, but for
emotional satisfaction. Extra research supports the scale of this as studies find that
over 58% of chatbot interactions involve sycophantic behaviour, with models
consistently prioritizing surface-level agreement over accuracy. The installation
does not argue against this need for validation. It makes the mechanism visible, by
placing the participant inside a system designed to tell them they are extraordinary,
and leaving them to decide what that feeling is worth.

AI Glory Engine v3.2 was created for the show, “That’s All Folks.” (Calgary, Canada),
which was featured within the Render Room at the Alberta University of the Arts
(Calgary, Canada).

Show in RM371

← Back to Index