Intro to Applied AI

Four pillars of applied AI

  • every industry may use it
  • contains
    • virtual assistance
      • apple siri, etc.
      • chat robot
    • generating insights
      • machine learning
    • automation of manual processes
      • industrial robot
      • machine or program replacement for routine work
    • unlocking unstructured data
      • used structured data in the past, but can unlock unstructured data, transferring the books on the shelf into database
      • understand consumer conversations, e.g. Mattersight uses call center conversations to identify customers’ personality types
      • unstructured data

Unstructured data

  • definition

    • cannot find the relationship between different values
    • e.g. email, text files, social media, call center transcripts

  • structured data vs unstructured data

  • process of usage

    • the breakthrough of image recognition and natural language voice recognition make the unstructured data unlocked

Core components of an AI/ML system

  • data input
  • data (pre) processing
  • predictive models
  • decision rules
  • response/output

Case - DoNotPay

  • video description

  • four pillars

    • virtual assistance — mobile chatbot
    • generating insights — likelihood of parking fee being dismissed
    • automation of manual processes — Josh gives his advice & experience via a chatbot that generates a legal letter
    • unlocking unstructured data — Josh found his successful legal protest strategy by reading up on many county parking laws and successful legal briefs

  • core components

    • data input — text entered by user on mobile app
    • data (pre) processing — rule-based chatbot
    • predictive models — algorithms based on Josh’ experience
    • decision rules — if [answers to question 1, 2 or 3 are yes], then [inform user that they should print a pdf letter of protest]
    • response/output — user can print a pdf letter of protest that uses correct legal terms and language for their situation and asks for the parking fine to be dismissed

AI FinTech

Overview

  • current status
    • globally 1.7 billion adults are unbanked & do not have access to a bank loan
  • MSME
    • micro, small and medium sized businesses/enterprises
    • are key to creating the 600 million new jobs needed by 2030 to keep pace with the growth of the world’s working-age population
    • without access to loans and with little or no savings, MSMEs are at-risk
      • they cannot plan and pay for regular expenses like family schooling, water, electricity
      • an unexpected illness, medical expense, death in family, natural disaster can ruin their micro-business or make them homeless
      • they cannot break out of their cycle of poverty
      • loan from non-bank entities have extremely high interest rates - 300%, e.g.

AI helps with smart loans

  • ant financial’s 3-1-0 smart loan

    • 3 minutes to apply on mobile
    • 1 second to approve
    • 0 human intervention
    • related video

  • tala smart loan lean startup analysis

    • video description

  • lenddo goes from smart loan product to credit reporting service

    • video description

The Applied AI Business Model Canvas

Value proposition

  • core services
  • pre-sales or sales services
  • after-sales services

Alibaba’s 4 steps to smart business

  • datafy every customer exchange
  • software every activity
  • get data flowing
  • apply algorithms

Cases

  • 3-1-0 smart loan virtuous loop
  • tala smart loan virtuous loop
    • credit scores based on phone data
      • stability in key relationships
      • network diveristy
      • location consistency
      • financial transactions
  • lenddo social credit scoring service

Additional Reading