Sid Mehta Works
AI Self

AI Self

Project details

Role

Solo

Year

2024-2025

Tools

Unity · ARKit · Figma · Google Cloud Vision

Client

Self Initiated

Project highlights

3

research arcs

AI · AR · VR

modes of exploration

Solo

research & prototyping

Overview

Ran a three-part research arc across AI, AR and VR, probing what machines actually perceive and designing an intelligence that lives alongside you rather than answers to you.

AI · AR · VRSolo · Self Initiated

Will Smith eating spaghetti, generated in 2023 and again in 2024

Google Cloud Vision run against my own photographs. My first experiment with machine perception.

Vision model output: objects, labels and faces scored against a photograph

A simple system to feed images through a model and watch what it decides is worth naming.

I asked the model about grief, loneliness, fear. It answered each with the cadence of weather.

The flip side. What if the coldness was solved? What if AI could help someone change?

The first UI concept. Pick a mascot. Talk to it. Over time it becomes a companion, not a tool.

Choose an AI you like: Meh the Cube, Fantastic Fox, or Baloo the Bear

Beyond the app. Not a screen you hold up, a layer that lives with you.

If you had to tell me something you are going through that you need help with, what would it be

AR walkthrough. Idea conception, not a live build.

Watch Chaise Walkthrough

I asked Claude if it dreams. This is that conversation.

Reference board: consumption, discoveries, nature, other AI, art and history

AI arrives. Not as a tool, as a presence trying to understand what it inherited.

Watch AI Self Experience

A different entry point. An AI assembling a sense of self from fragments of what it has learned.

AI dream of self, in four renders

Process sketches for how the scenes were structured.

Four more renders from the same series

Eva. The AI that wakes up. This is her opening.

Watch Full Experience
Reflection

The research started as a joke. Feeding images into Cloud Vision just to see how badly it misread them. But the patterns kept surfacing. Not about what AI fails at, but about what it has decided is worth noticing in the first place.

Chaise was the harder problem. If AI sees pattern instead of meaning, what would it feel like to have that living alongside you all day? The design question was always about restraint: when presence becomes intrusion, and how to build something that knows the difference. AI Self was the most difficult to make. Unity is unforgiving if you do not know what you are doing, and at the start I did not. What I ended up with was rougher than intended, but the roughness felt appropriate. Something trying to understand humanity by building a room out of it is going to get some of the proportions wrong.

Back to top
Next project Aananda Aananda explores how Jain philosophy reaches younger audiences without losing what make... Back to top