The document discusses the adoption of artificial intelligence (AI). It defines AI and different types of learning such as supervised, unsupervised, and reinforcement learning. Deep learning is described as a class of deep neural networks that has helped increase AI adoption. Examples of real-world AI applications are provided, such as predictive policing, self-driving cars, and AI assistants. Challenges to AI adoption include accessing data, defining business cases, and a lack of expertise. The document suggests when to apply AI, such as when human expertise is absent or the problem is too large for human capabilities.

1. Adoption of AI
The Great Opportunities for Everyone
Kan Ouivirach
Image Credit: BBC Focus - Science & Technology

2. Kan Ouivirach
Lead Software Architect

3. Outline
• What is AI?
• What Exactly AI is Doing: No Magic, Pure Code
• The Rise of AI Adoption
• Challenges and When to Apply AI
• Recap

4. Pioneer in AI
http://www-03.ibm.com/ibm/history/ibm100/us/en/icons/ibm700series/impacts/
On February 24, 1956, Arthur Samuel’s Checkers program

5. https://twitter.com/inteliment/status/933876632112136193
Impact of AI on Companies Today

6. —Andrew Ng
“AI is the new electricity.”
deeplearning.ai

8. http://www.deeplearningbook.org/

9. Strong AI vs. Weak AI

10. Strong AI vs. Weak AI

11. Strong AI vs. Weak AI

12. What Exactly is AI Doing?
https://www.pexels.com/photo/abstract-blackboard-bulb-chalk-355948/

13. Learning
Algorithm
Feature
Representation
Learning Process in General
https://www.instagram.com/p/Bb9nR3aB_u6/

14. Given x, y pairs, find a function f that will map
new x to a proper y.
y = f(x)
Supervised Learning

15. Supervised Learning: Neural Network
http://i-systems.github.io/HSE545/machine%20learning%20all/16%20Deep%20learning/CNN.html

16. Processor
Input 0
Input 1
Output
One or more inputs, a processor, and a single output
Perceptron
W 0
W 1

17. Processor
12
4
Output
0.5
-1
(12 x 0.5) + (4 x -1)
sign(2)
+1
Perceptron (cont.)

18. Input 0 Input 1 Label
12 4 -1
5 9 +1
Data:
Initial Weight:
Weight 0 Weight 1
0.5 0.5
We define:
Desired Guess Error
-1 -1 0
-1 +1 -2
+1 -1 2
+1 +1 0
Perceptron (cont.)

19. Processor
12
4
0.5
0.5
+1
(12 x 0.5) + (4 x 0.5) = 8
We then get:
New Weight 0 = 0.5 + (-2 x 12) = -23.5
New Weight 1 = 0.5 + (-2 x 4) = -7.5
Weight = Error x Input
New Weight = Weight + Weight
We actually need -1!
Input 0 Input 1 Label
12 4 -1
5 9 +1
Perceptron (cont.)

20. Processor
5
9
-23.5
-7.5
-1
(5 x -23.5) + (9 x -7.5) = -185
We then get:
New Weight 0 = -23.5 + (2 x 5) = -13.5
New Weight 1 = -7.5 + (2 x 9) = 10.5
Weight = Error x Input
New Weight = Weight + Weight
Perceptron (cont.)
We actually need +1!
Input 0 Input 1 Label
12 4 -1
5 9 +1

21. https://datasciencelab.wordpress.com/2014/01/10/machine-learning-classics-the-perceptron/
Perceptron (cont.)

22. https://trends.google.com/trends/explore?date=all&q=artificial%20intelligence
The Rise of AI Adoption

23. Traditional Machine Learning
http://i-systems.github.io/HSE545/machine%20learning%20all/16%20Deep%20learning/CNN.htmlDeep Learning for Computer Vision with Python by Dr. Adrian Rosebrock

24. Deep Learning
Deep Learning for Computer Vision with Python by Dr. Adrian Rosebrock http://i-systems.github.io/HSE545/machine%20learning%20all/16%20Deep%20learning/CNN.html
Automate!

25. Deep Learning for Computer Vision with Python by Dr. Adrian Rosebrock
Larger Network!
Deep Learning with Larger Network

26. https://en.wikipedia.org/wiki/Convolutional_neural_network
Convolution Neural Network
A Class of Deep Neural Networks

27. https://github.com/zkan/hello-keras
Convolution Neural Network (Demo)

28. In AI, generally think about 3
types of learning
• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning

29. 1 Most-Used Type of Learning
• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning

30. • Supervised Learning
• Unsupervised Learning
• Reinforcement Learning
• Transfer Learning
• Adversarial Learning
2 More Interesting Types of
Learning

31. Image Recognition
Transfer Learning

32. Image Recognition
Medical Diagnosis
Transfer Learning (cont.)

33. https://blog.statsbot.co/generative-adversarial-networks-gans-engine-and-applications-
f96291965b47
Adversarial Learning

34. Example of Adversarial Learning Application
https://arxiv.org/pdf/1512.00570v2.pdf
Attribute2Image: Conditional Image Generation from Visual Attributes by Yan et al.

35. Real-World Examples of
AI Applications
http://www.predpol.com/

36. Real-World Examples of
AI Applications
https://www.recode.net/2017/11/7/16614780/alphabet-driverless-cars-phoenix-arizona
http://www.bloomberg.com/news/articles/2016-07-19/google-cuts-its-giant-electricity-bill-with-deepmind-powered-ai
http://blog.fastforwardlabs.com/2016/04/11/new-tools-to-summarize-text.html

37. https://deepmind.com/research/alphago/
https://www.alphagomovie.com/
Real-World Examples of
AI Applications

38. More AIs in Various Industries
• Artificial intelligence has learned to spot suicidal tendencies from brain
scans (link)
• Microsoft’s AI is learning to write code by itself, not steal it (link, arXiv)
• Algorithm that can detect pneumonia from chest X-rays at a level
exceeding practicing radiologists (link)
• Neva automates customer service and support to deliver
unprecedented precision and quality. (link)
• AI can hunt down missile sites in China (link)
• The titans of AI are getting their work double-checked by students (link)

39. Challenges in AI Adoption
1. Accessing to data
2. No defined business case
3. Lack of people power
4. Lack of emotional intelligence
5. Better at specialized tasks
6. Difficult to collaborate between AIs

40. When to Apply AI?
• Human expertise is absent
• Humans are unable to explain their expertise (speech recognition,
vision, language)
• Solution changes with time (tracking, temperature control,
preferences)
• Solution needs to be adapted to particular cases (personalization)
• Problem is too big for our limited reasoning capabilities
(calculating webpage ranks, matching ads to facebook pages)

41. Recap

42. https://www.facebook.com/zuck/posts/10104242660091961

43. Are killer robots an imminent threat?
https://www.facebook.com/playgroundenglish/videos/566310110368990/

44. https://www.pexels.com/photo/action-android-device-electronics-595804/
Thank You~
Special Thanks to