AI2V Apply to get Value AI2V 101 By Alexandre Dietrich

AlexD.ai What is AI ?

AlexD.ai What is AI ?

Google search: “Artificial Intelligence definition” Encyclopaedia Brittanica: Dictionary: Artificial intelligence (AI), the ability of a digital computer the theory and development of computer systems or computer-controlled robot to perform tasks commonly able to perform tasks that normally require human associated with intelligent beings. The term is frequently intelligence, such as visual perception, speech applied to the project of developing systems endowed with recognition, decision-making, and translation the intellectual processes characteristic of humans, such as between languages. the ability to reason, discover meaning, generalize, or learn from past experience.

Stanford University – Computer Science Department: Q. What is artificial intelligence? A. It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.

Mckinsey & Company: Artificial intelligence: A definition AI is typically defined as the ability of a machine to perform cognitive functions we associate with human , such as perceiving, reasoning, learning, and problem solving. Examples of technologies that enable AI to solve business problems are robotics and autonomous vehicles, computer vision, language, virtual agents, and . AlexD.ai AI2V Function

AlexD.ai AI2V Function AI(S,S D)D à V

AI - Artificial Intelligence S - Situation D - Data V - Value à - Enablers / Inhibitors AlexD.ai Why not use Humans ? AI(S,(S, D) à V

AlexD.ai Why not use Humans ? (S,DD DD) à V AlexD.ai Why not use a Traditional Computer Program ? (S, D) à V

AlexD.ai +D = OK + D = OK AI? + = OK D AlexD.ai AI(S,1 D)1 à V 1 AI(S,2 D)2 à V 2 AlexD.ai AI(S,1 D)1 à V 1 AI(S,2 D)2 à V 2 AI(S,3 D)3 à V 3 AlexD.ai AI(S,2 D)2 à V 2 AI(S,3 D)3 V 3 … à AI(S,N D)N à V N AlexD.ai Present

Narrow AI(S, D) à V

AlexD.ai Future ?

General N N N

AI(S,1 D)1 à V 1

AlexD.ai AI Capabilities

AlexD.ai AI(S, D) à V

AlexD.ai Artificial Intelligence

Automated Reasoning Natural Language Knowledge Processing Representation

Computer Machine Learning Vision Robotics

Deep Learning

Neural Networks

Adapted from the book “Artificial Intelligence: A Modern Approach” – Stuart Russell and Peter Norvig. AlexD.ai Machine Learning

AlexD.ai Traditional Programming X Machine Learning Training Data Input Data Software Machine Learning Developer Learning Algorithm Engineer

Input Rules Data / Code

ML Model (Code)

Output

Output AlexD.ai Traditional Programming X Machine Learning

Smartphone Camera Web Page Field A = 123456

Accounts Database Pictures Database

B = 123456

Machine Learning output: if A == B: The probability that is equal is 87% . print(‘Account Number is correct’) else: print(‘Account Number is wrong’) if probability >= 85%: print(‘Welcome back’) else: print(‘Unknown Customer’)

AlexD.ai Heidenreich, H. 2018. The Future with Reinforcement Learning. Towards Data Science. AlexD.ai Deep Learning

AlexD.ai Multilayer Perceptron (Fully Connected Neural Network - FCNN)

1

X1 Y

X2

Weights

Adapted from University of Waterloo and University of Toronto AlexD.ai Recurrent Neural Network (RNN) Transformers (pre-trained models)

Convolutional Neural Networks (CNN)

BERT

Source: University of Toronto, Towards Data Science and AlexD.ai Computer Vision

AlexD.ai Fei-Fei Li, J. Johnson, s. Yeung, Convolutional Neural Networks for Visual Recognition, Spring 2018 AlexD.ai Facial Recognition Style Transfer

Augmented Reality

Generative Adversarial Network (GAN)

Source: Machine Learning Mastery, Microsoft, Medium and Realeyes. AlexD.ai Natural Language Processing

AlexD.ai Passage: “Toronto residents have been staying home and practicing social distancing for more than a month now, and the chilly spring weather has made doing so just a little bit easier.” BlogTO

Reading Comprehension: Model: NAQANet Question: Where are the Toronto residents? Answer: home

Semantic Role Labeling: Model: BERT

Sentiment Analysis: Model: RoBERTa Answer: Positive

Language Modeling: Model: GPT-2 Sentence: Toronto residents have been staying …

AI2 – Allen Institute for AI – Allen NLP - https://demo.allennlp.org AlexD.ai Language Translation Text Classification and Speech Recognition and Synthesis Named-Entity Recognition

AlexD.ai Situation

AlexD.ai AI(S, D) à V

AlexD.ai Workflow

Task Task Task Task Task Task 1 2 3 4 5 N …

Computer Computer Computer Program 1 Program 2 Program N … Human Human Human Human Human Role Role Role Role Role A B C D … N

AlexD.ai Input Data Action Outcomes AND / OR

Decision

Feedback

Inspired by the book 'Prediction Machines - The Simple Economics of Artificial Intelligence’ by Ajay Agrawal, Joshua Gans and Avi Goldfarb AlexD.ai Recognition

A B C 80% 72% 35% Input Data Action Outcomes

Recognition (CV, NLP, ML, DL)

Decision

Feedback

CV – Computer Vision NLP – Natural Language Processing ML – Machine Learning DL – Deep Learning AlexD.ai Prediction

87% Action Input Data Outcomes

à ? Prediction (ML, NLP, DL)

Decision

Feedback

AlexD.ai Discovery

Action Input Data + Outcomes

Discovery (ML, NLP, DL)

Decision

Feedback

AlexD.ai Generation

Generation Input Data (NLP, CV, DL) Outcomes

Decision Feedback

AlexD.ai Automation

Judgement (Algorithm)

87% Action Input Data Outcomes (Algorithm)

Prediction (ML, NLP, DL)

Decision

Feedback

AlexD.ai Human-Machine Interaction

Interaction

Judgement Generation (Algorithm) (NLP, CV, DL)

Output 87%

Input Data

Feedback Recognition/ Outcomes Prediction/ Discover (NLP, DL, ML, CV)

Decision

AlexD.ai Workflow

Task Task Task Task Task Task 1 2 3 4 5 N …

Computer Computer AI Computer Program 1 Program 2 Program N Gen … AI Hum Auto Human AI(HMI) Human Human AI Role Role AI AI Role AI Role HMI Pred Rec Dis A B D … N

Auto – Automation Dis – Discovery Gen – Generation HMI – Human-Machine Interaction Pred – Prediction Rec – Recognition

AlexD.ai Task N …

Automation Discovery Task Generation 3 Human-Machine Workflow Interaction Prediction Recognition Task S 2

Task 1

AlexD.ai Data

AlexD.ai AI(S, D) à V

AlexD.ai AI(S, D) à V

AlexD.ai Structured Data Unstructured Data Small Data Big Data AI(S, D) à V Messy Data Training Data Input Data Feedback Data

Data Management AlexD.ai Enablers and Inhibitors

AlexD.ai AI(S, D) à V

AlexD.ai AI(S, D) à V People Tools Infrastructure Process

AlexD.ai AI(S, D) à V Transparency Explainability Ethics

AlexD.ai People

AlexD.ai Data Scientist

Machine Learning Business Professional Data Engineer Software Developers Business Professional Engineer Business Analyst Data Analyst Software Engineers Business Analyst Data Analyst Machine Learning Engineer

AI(S, D) à V

AlexD.ai Value

AlexD.ai AI(S, D) à V

AlexD.ai Get value from:

Human augmentation

Task automation

Products and services innovation

Lifelong learning

Just for fun

AlexD.ai Source: McKinsey & Company, 2019 AlexD.ai Source: McKinsey & Company, 2019 AlexD.ai Conclusion

AlexD.ai AI(S, D) à V

AlexD.ai AlexD.ai Thank You

Alexandre Dietrich Artificial Intelligence Enthusiast

Linkedin: https://br.linkedin.com/in/alexdiet Twitter: @alexdiet42 Youtube: youtube.com/c/AlexandreDietrich

AlexD.ai