Happy Path Programming #109 AI, LLMs, and MCP with Lize Raes
Feb 21, 2025
Lize Raes, an AI and LLM expert, dives into the mechanics of language models and their transformative capabilities. The discussion reveals how training data and model architectures enhance AI performance. Lize also unpacks exciting advancements in protein folding technology and its impact on drug discovery. The integration of AI into user interfaces is explored, highlighting tools like LangChain4J and the Model Control Protocol (MCP). Finally, the conversation touches on AI's role across industries, emphasizing its potential to augment human creativity and productivity.
AI Snips
Chapters
Transcript
Episode notes
Lize's Path to Software
- Lize stumbled into software development after studying electro-technology and finding hardware debugging too difficult.
- Her final thesis project, a breathing sensor, involved a fragile glass fiber that broke after finally working.
Hardware Background Informs Software
- Lize's background in hardware influences her software development approach, making her conscious of resource usage.
- She was initially shocked by Java's resource consumption compared to C++, where optimization was crucial.
LLMs as Equations
- Large language models (LLMs) function like complex equations with countless terms, relating input data to outputs.
- While seemingly magical, these models use layers of interconnected nodes with weights determining how information propagates.