Skip to main content

GenAI and LLMs

Generative AI​

Generative AI is a subset of Deep learning that uses models to perform human-like tasks.

LLMs​

LLMs (Large Language Models) are a type of generative AI model that uses transformers to generate text.

Prompts and Completions:

  • Prompts are the input to the model.
  • Completions are the output of the model.
  • For example, if the prompt is "Hello, how are you?", the completion is "I am fine, thank you!".
  • The act of generating text using models is called inference.

Use Cases:

  • Text Generation: Generating text based on a given prompt.
  • Text Completion: Completing the text based on a given prompt.
  • Text Summarization: Summarizing a text.
  • Text Translation: Translating a text from one language to another. Or Translating natural language to code.
  • Text Question Answering: Answering questions based on a given text.
  • Entity Extraction: Extracting entities from a text.
  • Flight Information: Connecting the LLM to an external source.

Generative AI Project Lifecycle​

  1. Scope: Define the scope/user case of the project. Generation, Named Entity Recognition, etc.
  2. Select: Start with an existing model or train a new model.
  3. Adapt and Align Model
    • Prompt Engineering: Design the prompt to get the best out of the model. In context learning, few shot, zero shot, etc.
    • Fine Tuning: Fine tune the model on a specific dataset.
    • Align with Human Feedback: Align the model with human feedback.
    • Evaluate: Evaluate the model performance. The process is highly iterative.
  4. Application Integration: Connect the model to the application.
    • Optimize and deploy model for inference: Optimize the model for deployment.
    • Augment model and build LLM-Powered Application: Augment the model and build an LLM-Powered Application. Reduce the tendency to hallucinate.