Fake News: Can Artificial Intelligence Help?

Keynote lecture at the workshop on "Studying Russian Disinformation and its Effects in Europe: Theoretical, Methodological and Practical Challenges" organised by the Graduate School for Eastern and South Eastern European Studies at Ludwig-Maximilian University of Munich, Germany.

The event is an interdisciplinary workshop that focuses on the analysis of the attempts to spread false or misleading information in Europe. Renowned experts from Germany and abroad discuss the forms and consequences of current disinformation efforts as attributed to Russian authors in particular. Particular attention is paid to the theoretical, methodological and practical challenges of research on such phenomena.

Workshop web page (English): http://www.gs-oses.de/event-detail-317/events/workshop-munich-studying-russian-disinformation-and-its-effects-in-europe-theoretical-methodological-and-practical-challenges.html
Workshop web page (German): http://www.gs-oses.de/event-detail/events/workshop-muenchen-studying-russian-disinformation-and-its-effects.html

Transcript

1. Fake News: can Artificial Intelligence help? Marco Bonzanini Workshop: Studying Russian disinformation and its effects in Europe Ludwig-Maximilian University of Munich, Germany
2. — Betteridge’s Law of Headlines “Any headline 
 that ends in a question mark 
 can be answered 
 by the word no”
3. ARTIFICIAL INTELLIGENCE
4. AI
5. AI Computer Science Mathematics
6. AI Computer Science Mathematics Linguistics
7. AI Computer Science Mathematics Philosophy Psychology Linguistics Neuroscience
8. AI Computer Science Mathematics Philosophy Psychology Linguistics Neuroscience
9. Intelligent Agents Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
10. Intelligent Agents Any device that: Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
11. Intelligent Agents Any device that: • perceives its environment Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
12. Intelligent Agents Any device that: • perceives its environment • takes action Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
13. Intelligent Agents Any device that: • perceives its environment • takes action • maximises the chances of success Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
14. Intelligent Agents Any device that: • perceives its environment • takes action • maximises the chances of success Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
15. Intelligent Agents Any device that: • perceives its environment • takes action • maximises the chances of success Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003 what’s the goal?
16. Intelligent Agents Systems that: Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
17. Intelligent Agents Systems that: • act like humans Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
18. Intelligent Agents Systems that: • act like humans • think like humans Russell and Norvig, Artificial Intelligence: A Modern Approach, 2003
19. Act Like a Human
20. Act Like a Human A. Turing (1950) Computing Machinery and Intelligence
21. Act Like a Human A. Turing (1950) Computing Machinery and Intelligence ??? Human AI
22. Act Like a Human A. Turing (1950) Computing Machinery and Intelligence ??? Human AI a.k.a. The Turing Test
23. Passing the Turing Test
24. Passing the Turing Test Ex Machina, 2015
25. AI: Main Ingredients
26. AI: Main Ingredients AI {
27. AI: Main Ingredients • knowledge AI {
28. AI: Main Ingredients • knowledge • reasoning AI {
29. AI: Main Ingredients • knowledge • reasoning • language understanding AI {
30. AI: Main Ingredients • knowledge • reasoning • language understanding • learning AI {
31. Machine Learning
32. Machine Learning ML gives computers the ability to learn
 without being explicitly programmed
33. Types of Machine Learning
34. Types of Machine Learning • Supervised • Unsupervised • Semi-supervised • Reinforcement
35. Types of Machine Learning • Supervised • Unsupervised • Semi-supervised • Reinforcement Sample inputs + Desired outputs
36. Types of Machine Learning • Supervised • Unsupervised • Semi-supervised • Reinforcement Find structure in sample input
37. Types of Machine Learning • Supervised • Unsupervised • Semi-supervised • Reinforcement Incomplete training signal
38. Types of Machine Learning • Supervised • Unsupervised • Semi-supervised • Reinforcement Reward or punishment
39. Applications
40. Applications • Filter spam emails
41. Applications • Filter spam emails • Play poker/chess/go
42. Applications • Filter spam emails • Play poker/chess/go • Drive safely on a tortuous track
43. Applications • Filter spam emails • Play poker/chess/go • Drive safely on a tortuous track • Drive safely on a busy road
44. Applications • Filter spam emails • Play poker/chess/go • Drive safely on a tortuous track • Drive safely on a busy road • Answer simple factual questions
45. Applications • Filter spam emails • Play poker/chess/go • Drive safely on a tortuous track • Drive safely on a busy road • Answer simple factual questions • Give legal advice in a specialised field
46. Applications • Filter spam emails • Play poker/chess/go • Drive safely on a tortuous track • Drive safely on a busy road • Answer simple factual questions • Give legal advice in a specialised field • Translate spoken Eng-to-Swe real-time
47. Applications • Filter spam emails • Play poker/chess/go • Drive safely on a tortuous track • Drive safely on a busy road • Answer simple factual questions • Give legal advice in a specialised field • Translate spoken Eng-to-Swe real-time • Have a pleasant 1-hour conversation
48. http://www.independent.co.uk/life-style/gadgets-and-tech/news/tay-tweets-microsoft-creates-bizarre-twitter-robot-for-people-to-chat-to-a6947806.html
49. http://www.stuff.co.nz/technology/digital-living/78265631/Microsofts-AI-teen-turns-into-Hitler-loving-Trump-fan-thanks-to-the-internet
50. AI in Summary
51. AI in Summary • What are you trying to solve?
52. AI in Summary • What are you trying to solve? • Current state of the art: narrow problems
53. AI in Summary • What are you trying to solve? • Current state of the art: narrow problems • Generalising is difficult
54. AI in Summary • What are you trying to solve? • Current state of the art: narrow problems • Generalising is difficult • Artificial General Intelligence is not here
55. AI in Summary • What are you trying to solve? • Current state of the art: narrow problems • Generalising is difficult • Artificial General Intelligence is not here (yet )
56. FAKE NEWS
57. Google Trends for “fake news”
58. Google Trends for “fake news” No fake news before November 2016?
59. http://www.bbc.co.uk/news/uk-40098804
60. https://www.theguardian.com/media/2017/jun/13/fake-news-manipulate-elections-paid-propaganda
61. https://www.theguardian.com/media/2017/jun/13/fake-news-manipulate-elections-paid-propaganda ~$55k to discredit a journalist ~$200k to instigate a street protest
62. https://twitter.com/carolecadwalla/status/872767628712005632
63. https://twitter.com/carolecadwalla/status/872767628712005632 Democracy
 to the highest bid?
64. News (Fake or Not)
65. News (Fake or Not) • Easy to manufacture
66. News (Fake or Not) • Easy to manufacture • Difficult to verify
67. News (Fake or Not) • Easy to manufacture • Difficult to verify • 83.1M posts/month on WordPress.com
68. News (Fake or Not) • Easy to manufacture • Difficult to verify • 83.1M posts/month on WordPress.com • 500M tweets/day on Twitter
69. Information Overload
70. AI vs FAKE NEWS
71. Information Overload
72. Information Overload Recommender systems can choose for us!
76. Filter Bubble
77. Filter Bubble stream of news recommender
 system environmentalists football nuts
78. Filter Bubble news on
 climate change
79. Filter Bubble are you going
 to “like” it?
80. Filter Bubble
81. Filter Bubble transfer market news
82. Filter Bubble are you going
 to “like” it?
83. Filter Bubble
84. https://www.theguardian.com/technology/2017/may/22/social-media-election-facebook-filter-bubbles
85. Fact Checking
86. Fact Checking Full Fact, The State of Automatic Factchecking, 2016. 1. Monitor
87. Fact Checking Full Fact, The State of Automatic Factchecking, 2016. 1. Monitor 2. Spot claims
88. Fact Checking Full Fact, The State of Automatic Factchecking, 2016. 1. Monitor 2. Spot claims 3. Check claims
89. Fact Checking Full Fact, The State of Automatic Factchecking, 2016. 1. Monitor 2. Spot claims 3. Check claims 4. Create and publish
90. Monitor
91. Monitor • What to monitor?
92. Monitor • What to monitor? • How to access?
93. Monitor • What to monitor? • How to access? • Scalability & robustness
94. Monitor • What to monitor? • How to access? • Scalability & robustness • Long-term availability
95. Monitor • What to monitor? • How to access? • Scalability & robustness • Long-term availability • Engineering problems, not really AI
96. Spot Claims
97. Spot Claims • Human judges don’t always agree 
 
 

98. Spot Claims • Human judges don’t always agree • Text analysis to the rescue 
 

99. Spot Claims • Human judges don’t always agree • Text analysis to the rescue • Natural language is “difficult”
 (paraphrasing, common sense, idiomatic expressions, …)
100. Check Claims
101. Check Claims • Look-up high-quality knowledge base
102. Check Claims • Look-up high-quality knowledge base • … how do you model the universe?
103. Check Claims • Look-up high-quality knowledge base • … how do you model the universe? • Context is important
104. Create and Publish
105. Create and Publish • Communicating the results is crucial
106. Create and Publish • Communicating the results is crucial • Automated journalism?
107. Create and Publish • Communicating the results is crucial • Automated journalism? • How do you build trust?
108. — DJ Patil “The hardest part of data science is getting good, clean data. Cleaning data is often 80% of the work” “We didn’t have better algorithms, we just had better data.” — Peter Norvig
109. Can AI help? Is this the right question? Automation vs AI Data Quality Summary