Generative AI 101 Slides

Agenda - Overview, history, current state, whats coming, limitations and mitigations
Slide 1 for generative AI 101
Slide 2 for generative AI 101
Slide 3 for generative AI 101

Learn how software evolved from coding to data driven ML. Now genertive AI is even generating data and code. This will change product business model and also product management.

Foundational model and new product

New products will be built around foundational model Workflow and propriety dataset are key for enterprises
Slide 4 for generative AI
Slide 5 for generative AI 101
Slide 6 for generative AI

Foundations models are train on broad and generic data. One can use propriety data and use these models for specific tasks in specific domain e.g. finance, health New businesses can be built by training model on top of foundation model. Key is propriety data, domain knowledge and workflow.

Generative AI applications

Writing speech, content, job description, automation of customer support, sales and marketing emails, image and logo generation
Slide 7 for generative AI
Slide 8 for generative AI 101
Slide 9 for generative AI 101

Many verticals are impacted because of progress in generative AI. This is just a start and more advance applications will evolve. Applications specific to propriety dataset for doing a narrow task will also evolve. More importantly generative AAI can solve cold start problem by generating dataset for model training.

Generative AI Applications and Finance

Applications in finance involve customer support, analyzing earnings call build data products for multiple use cases
Slide 10 for generative AI
Slide 11 for generative AI 101
Slide 12

Finance can benefit from applications that help investors, portfolio managers.

  • There is significant mount of text data and it need to be analyzed with numbers. One can generate summary from document, earnings calls.
  • Using generative AI one can convert natural language to queries that understand finance concept and it can provide assistance to portfolio managers.
  • Customer support, collections can use generative AI to understand customer complaint, intent and provide response and follow up for collection.
  • Significant amount of document processing is require to comprehend document, validate info and prepare memo for underwriting.
    • Generative AI can do this provided it is trained for specific tasks.

    Tech stack and Modeling Architecture

    There are many architectures - Transformer Auto encoder, GAN, Diffusion model and RL
    Slide 13 for generative AI
    Slide 14 for generative AI 101
    Slide 15 for generative AI

    Generative AI is based on "comprehend existing" data and determine trajectories data can take. It uses it for geenration

  • Diffusion architecture is suitable for generation
  • Transformer are suitable for language gentration in sequence