Notes on AI

Random notes from basic learnings;

• Foundation models are the basis of every generative AI model and can be fine-tuned for specific tasks or fields

  • eg: NPL: CHATBOTS
  • Computer Vision: Image generation (e.g., DALL·E)

• There are diff types of foundation models: (There are a lot more)

  • LLMs
  • DIffusion Models:
    • They generate data that is similar to the data they’ve been trained on.

Embeddings and Vector Databases

• Embeddings of your knowledge base (docs,faqs) are created using an embedding model like OpenAI’s text-embedding-ada-020.
• Embeddings get stored in vector databases like Pinecone.
• When user asks a question, query gets converted into an embedding, then perform a search in the vectorDB for similar embeddings, and retrieve relevant content.
• Use an LLM to generate a response based on the retrieved content

##### Example usage 💡
Medical Q&A System:
 
1- User asks: "What are symptoms of diabetes?"
2- Question gets converted to embedding → finds relevant medical documents
3- Fine-tuned medical model gets these documents and generates response using proper medical terminology and formatting

• Embeddings as a smart librarian who knows where to find relevant books
• Fine-tuning as teaching the librarian how to explain those books in a specific way