Inside the Black Box: Cracking AI and Deep Learning
How do Large Language Models like ChatGPT work, anyway?
Episodes (36)
Three Ways to Be Wrong About the Past Tense
Not at This Address
My Advisors Argued This for Thirty Years. Now You Can Check
Inside the Past Tense: Raw Letters, Raw Sound
Forcing a Neural Network to Add -ed
The Wug Test for AI
Learning the Past Tense in AI
Fine Tuning Lora: It's Not What You Think
When Fluent Answers Start Sounding True
Why Your Brain Believes the Model
When Polished Answers Feel Finished
What Seneca Teaches Us that Marcus Couldn't
The Pattern Holds for Another Author
The Pattern Holds
Cracking Open the Black Box
Inside a Fine-Tuned Language Model
What Counts as Structure? From Harris and Elman to Today’s Neural Nets
Building a House Without Blueprints: When Interpretability Tools Work — and When They Don’t
I Told My LLM Not to Say "Empower"
Beyond the Surface of AI Intelligence
Unlocking BERTs Hidden Grammar
Cracking the Code of AI Interpretation
Decoding GPTs Hidden Circuits
Decoding Attention and Emergence in AI
When Knowledge Battles Noise in GPT Models
Inside Circuits: How Large Language Models Understand
Hallucinations, Interpretability, and the Seahorse Mirage
How Transformers Stack Meaning Like Finnish Words
The Mandela Effect in AI: Why Language Models Misremember
Bridging Circuits and Concepts in Large Language Models
How Transformers Turn Words Into Meaning
Can Smaller Language Models Be Smarter?
The Weird Geometry That Makes AI Think
Can We Fix It?
Using Symbolic AI to Explain LLMs
