This is an opening presentation from RIGA COMM 2019 Human + Intelligent Machine track. https://rigacomm.com/en/program/#2 How can we use intelligent machines in business workflows? There are 3 main things to remember - learning, teaching and understand environment. We will have humans in the loop.

3. RIGA COMM 2017, 2018

4. 4

5. Data visualization Community

6. Local AI, Machine Learning Community

8. 8

9. Reading makes a country great
9

10. What to expect this year
10

11. 11
If it could be
predicted….
What could be
expected result?
What could be
impact on
organization?
What kind of
data need to be
collected?
Prediction Impact Impact on strategy Information
Business need first!

12. Business Decision Automation
Learning, Teaching, understand
environment
Aldis Erglis

13. Why we don’t have self-driving cars yet?
13

14. The answer – Slow learning!
In Predictive environment you
can predict next stages
80 Hours of training required for
best machines to reach level
every human gets in 15 minutes
14
Yann LeCun
Atari game
Starcraft
For best models it’s
require 200 Years of
training to play better
than human in single
map, single type of
player

15. Get back to self-driving cars with best
reinforcement learning algorithms available
• 20 hours for human, mostly without
accidents
15
• Million hour of training
• Thousands killed pedestrians
• Hit thousands of trees
• Run from cliffs several time
We have predictive models that can predict uncertainty

16. How to learn model of the world?
16

17. If you think about it,
business processes are
the result of a series of
decisions.
17
Organizational structure
BI reports, KPI
libraries
Strategy, goals
Business Processes
Mapping
Mapping
Mapping
Mapping
Goal 1
Goal 2
CEO
CFO COO CIO
Model of an organization

18. Andrew Ng
If a typical person can do a mental
task with less than one second of
thought, we can probably automate
it using AI either now or in the near
future.
18
< 5 Sec < 7 Sec

19. The five steps to automating decisions
19
Step Key action
Datafy the physical world Put capabilities and assets online
Software the business Encode decision-making chains
Get data flowing Institute APIs to allow data
connections
Record data in full Record “live data” in its entirety
Apply machine-learning algorithms Coordinate and optimize
Assess readiness of your enterprise for automating decisions

20. As humans we suffer from biases
20 https://hbr.org/2016/10/noise
Daniel Kahneman
Organizations expect consistency from professionals:
Identical cases should be treated similarly, if not identically
Nobel Prize
in Economics

21. Why we want to automate decisions?
We think it is possible….
21

22. 22

23. Xrd Technology Revolution
Information revolution – now?
No, it is done by Gutenberg
in 15th century
Martin Luther real beneficiary
Next was telegraph and
railroad
23
Former FCC chairman Tom Wheeler discusses his
book, "From Gutenberg to Google", at Politics and
Prose on 2/10/19.

24. What is business decision automation
24
Copy data
Enter data
Sort documents
Prioritize work
Do Analysis
Make Decisions
The Knowledge Worker Spectrum
Worker 1Worker 2
Automated
operations
Automated
decisions
Translate
Classify Documents

25. Human + Machine (Narrow)

26. How narrow AI is?
26

27. Narrow AI/ML
We can’t trust AI systems
built on deep learning alone
27
Gary Marcus
Geoffrey Hinton
DL pioneer, Geoffrey
Hinton: “My view is throw
it all away and start
again”

28. Limitations of current technology
Yann LeCun
AI Director at Facebook
Small Data
Causality

29. Three waves of AI
29

30. Cognitive architecture
30
Perception
(input)
Action
(output)
API, Text, Image, Voice Text, Image, Voice
Short-Term memory
Analytical Path
(Directed graph)
Context
(List)
Learning
Decision
Procedures
(alghoritms)
Long-term memory
Procedural Semantic
Cases
(Sythetic)
Cases
(Sythetic)
Influence
graph
Topics
AP to Cases
transformation
Inference
engine
Processing (Parser)

31. AI vs GAI vs Human
31
Actually very specialized
Actually very specialized
Does not exit

32. SMALL DATA

33. Machine Teaching
33

34. 34

35. CAUSALITY

36. Causality
36
Development of Western science is based
on two great achievements: the invention
of the formal logical system (in Euclidean
geometry) by the Greek philosophers, and
the discovery of the possibility to find
out causal relationships by systematic
experiment (during the Renaissance). Albert Einstein
(1953)

37. How to explain population changes
37

38. 38

39. Causality to describe environment
39

40. Why employees learn
professional skills?
40

41. Future of automated factory
41

42. Future of decision automation
42
Machine Teaching
Human + Machine
Motivation
& Incentives