Research
Our research explores the mathematical, structural, and economic dimensions of AI systems — ensuring that as intelligence scales, its connection to human creativity, attribution, and insight is preserved.
How a Transformer Model Loses Attribution: A Step-by-Step Example
As transformers grow more capable of understanding, they grow less capable of remembering who taught them.
Why AI Isn't Built for Attribution: Understanding Transformers
Transformer AI models excel at language generation but can't track sources—resulting in fabricated citations and misattributed quotes. This research explores why attribution fails in transformers, examines real-world consequences, and reviews emerging solutions for the provenance problem.