EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·Aug 24·5 min read
Agent Is Not the Model
TL;DR
Understanding the distinction between agents and models is crucial in AI discussions. An agent system comprises a model, an inference service, and a harness for user interaction.
✦ Why It Matters
Clarify your terminology when discussing AI to improve collaboration and understanding.
Key Takeaways
- Differentiate between models, harnesses, and inference services in your projects.
- Use precise terminology to improve communication and troubleshooting.
- Evaluate how your harness shapes model outputs for better performance.
- Identify and address issues by naming the specific layer involved.
How It Works
An agent system integrates a harness that processes inputs and outputs, an inference service that executes the model, and the model itself, which performs the core computations. This structure allows developers to customize interactions and optimize performance based on the specific needs of their applications.