I started my dive into AI in 2008 writing a Boid / Crowd system for my thesis while in art college, School of Visual Arts.
It was an insane particle script + 3d animation cycles in Maya haha.
Then I did Boid movement, navigation, & obstacle detection in animated films for 5 years at Blue Sky Studios, using Houdini.
I dove into Style-Transfer AI & Long Short-Term Memory (LSTM) training in 2019-2020,
Like making a Node.js server (web site) understand my voice & auto google search for me.
Since then, I've been developing different multi-media AI structures in my spare time.
In 2015 I decided I'd cram a machine learning AI into a single-board computer, a Jetson TK1, by the end of 2026.
Something that could write down what I say,
Use vision to understand an object simply went out of frame.
Yet "knows" if it looks over, the object is still there; 'Attention'
At the end of 2023, this evolved into a deep learning AI crammed into, likely, a Jetson Nano.
As something to infer what I mean, from what I say,
Or give a "thought" on what it saw or heard in the world around it.
'Machine Learning' is AI that can learn basic patterns.
'Deep Learning' is Machine Learning,
But uses neural networks to form patterns of patterns.
Realistically, I'd just be happy to make something that can understand what I say and can give a semi coherent response without an internet connection.
As of May 24th 2025, I've started on the core of the AI,
But still testing different structure's ability in adapting to stimuli.
... It really seems like any network could work for most things, but some are better than others per task.
You could guess,
All the recent AI hullabaloo (2019-...)
Has been quite serendipitous for my creation!
I made a couple GNN sctructures to help show how GNNs and MPNNs work visually,
Mostly just to help me understand them.
GNNs are Graph Neural Networks,
And MPNNs are Message Passing Neural Networks.
Both are types of neural networks,
which consider the relationships between data points.
But MPNNs send messages between connected node neighbors-of-neighbors to update node states.
They are AI networks that can be used to predict connections between ideas / things / data in a system. While looking into Peter Velickovic's work on GNNs,
Saw a paper on mimicing a Dyjkstra Path Finding algorithm using a GNN.
The first video shown here is a test of that paper,
With my own implementation.
Then decided it would be best to show a better use case for GNNs.
To show off their capibilities in what they're best at.
So I set up an 'online matchmaking' system,
Showing neurons send multiple messages between connected nodes.
Here the nodes have dashed orange lines to show the 'potential teammates' between nodes.
Multiple messages get sent between neighbors before the final 'prediction' is made.
And the 'prediction' is the final team of 4 players that should help build a balanced team, connected in Red.
It asks all of the connected nodes for their 'player affinity' for a new game,
Having individual stats 'shift' per game to mimic a player acting differently per game.
Then to use those stats to find 'potential teammates' for a balanced team.
Example Player Stats:
- Name: 'P1' or 'Player 1'
- Offense: 0.80
- Support: 0.40
- Tank: 0.20
- Engineer: 0.65
The values will be used to find other players that have supportive traits,
Helping to construct a team of players that can work together.
This becomes the core influence on the messages sent between nodes.
Which contains 'Team Affinity', 'Role Balance', 'Skill Match' and 'Selection Confidence'.
These 4 values are what's sent between nodes to help determine if a player is 'willing' to change teams,
Influenced by the neighbor-of-neighbor player's stats.
It's pretty interesting to me,
Seeing these changes in potential teammates as the game rounds progress.
As the players change their stats, the potential teammates change between game rounds.
I'd like to imaginge some multiplayer games are using these type of statistics to help build teams.
Most games have "stat screens" that show off a player's stats,
But I wonder if they use these stats to help build teams.
Most seem to go off XP,
But I'm sure the systems are more complex than that.