#Azure #MicrosoftAIJourney Feedback Forms

Transcription

1 #Azure #MicrosoftAIJourney Feedback Forms

2

3 21 st September, th October, th October th November, th November, nd November, th November, th December 2018 What AI means to Microsoft and Microsoft partners AI without a PhD - Exploring speech, text, vision and bots HOL - Create a Cognitive Search solution for Enterprise Documents Getting to grips with AI and Machine Learning HOL - Predictive Maintenance AI - from theory to production Deep into data science and AI AI Conference Day Feedback Forms

4

5

6

7

8

9 Artificial Intelligence Historical Definition Weak AI Algorithmic data science and machine learning that provides foundational capabilities to machines without explicit programming Strong AI A combination of weak AI capabilities that approximate or exceed human capabilities (aka, Artificial General Intelligence) INCEPTION OF AI AI WINTER AI SPRING AI SUMMER 1956 Computer researchers envisioned a machine that will eventually be as intelligent as a human 1973 As realization of AI waned, investment disappeared 2012 Algorithmic and statistical machine learning + cloud triggered AI growth 2018 Democratized AI as APIs, vision, hearing, reading, translation, learning and AI everywhere

10 Challenges with AI Distorted Views of AI Hollywood Charlatans Sales pitches and promises Bias, Et Cetera Bias in data sets can have individual and societal impacts The system can be designed with or learn bias over time Infallibility Expectation The system can t make a mistake If mistakes are made, then the system is unreliable Ethical Challenges We CAN us AI, but SHOULD we for purpose XYZ? May start as innocent, but quickly evolve to ethical quandary Neutrality Perception The system isn t human so it is neutral The system is an impartial, unbiased advisor or judge AI Model Flexibility Static models can experience concept drift Models over time may become less accurate and decay As implementers of AI, we need to address and work to resolve these challenges

11 Insights everywhere Organisations are collecting data and telemetry to drive new insights Open Source Python and R are the AI languages that are winning the race Both are VIP technologies in the Microsoft stack Beginning of technology maturity = Devops Culture, Processes and Tools TDSP and Azure ML Services Organisations are now AI First Using AI to enable digital transformation and differentiate from the competition. (as opposed to price wars) Cloud first and Serverless architecture Training and Operationalizing AI with containers and clusters

12 AI Momentum Number of AI scientific papers on arxiv doubling every 18 months AI languages are gaining more market share

13

14

15

16

17 Managing Managing machines and datasets What are the traditional AI pain points Tracking and replicating Tracking and replicating experiments Sharing Sharing Results seamlessly

18 AI Development Database Development Development

19 Create/retrain model Monitor Register model Deploy image Cloud Light Edge Heavy Edge Create scoring files and dependencies Create and register image

20

21 Seductive Siris Self-Replicating AI The Singularity Rampaging Robots In 2013 s Her, Joaquin Phoenix falls in love with an intelligent operating system named Samantha. In 1995 s Screamers, scientists create a self-replicating weapon with one purpose: to destroy all life. AI vies to take over the world in 1982 s classic Tron. In 1973 s Westworld, recently reenvisioned as a different TV series by HBO, murderous androids go on a killing spree in a futuristic Disneystyle theme park. Feeling Machines Androids Among Us Human Enslavement Mind Upload In the 1999 movie Bicentennial Man, a household robot experiences emotions, creative thoughts, and eventually develops sentience. Artificial beings infiltrate society undetected in TV series Battlestar Galactica. In the 1999 movie The Matrix, all life on Earth is an elaborate facade. The robots are really the ones in command, but you wouldn t know it until you take the red pill. Digitized humans gain immortality and then wreak havoc, such as in 2014 s Transcendence.

22

23

24

25

26

27

28

29 Region Economic Impact of AI (2030) % of Total China $7.0 trillion 44.6% North America $3.7 trillion 23.6% Northern Europe $1.8 trillion 11.5% Developed Asia $0.9 trillion 5.7% Southern Europe $0.7 trillion 4.5% Latin America $0.5 trillion 3.2% Rest of World $1.2 trillion 7.6% Total $15.7 trillion 100.0% Industry 2035 GVA (Baseline) 2035 GVA (AI steady state) Manufacturing $8.4 trillion $12.2 trillion Professional Services $7.5 trillion $9.3 trillion Wholesale & Retail $6.2 trillion $8.4 trillion Public Services $4.0 trillion $4.9 trillion Information & Communication $3.7 trillion $4.7 trillion Financial Services $3.4 trillion $4.6 trillion Construction $2.8 trillion $3.3 trillion Transportation & Storage $2.1 trillion $2.9 trillion Data from PwC and Accenture

30

31

32

33

34

35

36

37 Ubiquitous Connectivity AI Big Data

38 Bottleneck 54 regions worldwide 140 available in 140 countries

39

40

41

42

43

44

45 AI Highly Technical Team Huge Market Exciting use case or common good

46 This material is provided for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED.