# Generative AI
AI -> ML -> DL(deep) -> Generative AI (subsets)
That focuses on generative content (text, image, audio, code etc)
Generative Model -> a branch of ML which mathematically approx the world object
Factors making Generative Model possible now
* Large Datasets
* availble of computation resources
* innovative DL models
Generative AI use cases
* synthetic data generation
LLM and Generative AI
Generative AI -> LLM + Foundation Models (GPT-4 etc)
How Do LLM work
encoder, decoder, transformer
Common LLM
Dolly -> Databricks (march 27, 2024)
Common LLM task
* content creation
* summarization
* question answer
* machine translation
LLM business use case
* customer persoanlization & segmentation
* feedback analysis
* virtual assistants
* code genration
LLM Application
LLM Falvours -> open source models (metabricks, llama fb) + proprietary models or LLM as service (openai, anthropic claude)
choose right model flavor ( LLM model selection factors -> consider privacy, quality, cost, latency)
Proprietary model (pros - quality, speed, cons -> cost, data privacy, vendor lock in)
open source models (pros - task tailoring, inference cost, control, cons - upfornt time investment, data requirment, skill sets)
Pre-trained model and training - (initial training with large corpus of data) - Foundation model + our data with fine tuning
FIne tuning model -> pre trained model/Foundation model + retrian on small data sets
Eg question answering model -> FOundation model + question anser data training
Mixing LLM flavors /models in a workflow (langchain helps on this) (multi-LLM)
eg summary + sentiment LLM
RAG - retrieval augmented Generation (enhance LLM output with external datasource)
Issue - new data comes in -> requires retriaing the model
Solution - RAG -> input — RAG + external vector database — return releavnt inputs to trained model
Databricks AI + mossaic ML
Industry impacted with Generative AI
tech, bank, pharma
customize + secure env + our data
how to prepare for AI revolution -> act urgently, underatand AI, develop strategy, identify use case, invest
potential risk and challenges
legal, ethical (bias, misinformation), social, environmental, security, intellectual property, issue
dont have forgetting feature in models, can reproduce model data,
Active Regulatory Area
* EU AI Act
Bias reinforcement loops (bias input and output)
hallucination (intrincic and extrinsic
how to address each issue like ethical issue, privacy issue ec
AUdit -> model, governance, application
Human - AI interactio and impact on society
---
# Agent Instructions: Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter:
```
GET https://www.pranaypourkar.co.in/the-programmers-guide/ai/generative-ai.md?ask=
```
The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.