The document discusses the progress and challenges of artificial intelligence (AI), highlighting advancements in perception, reasoning, and collaborative systems. It addresses the risks associated with autonomous AI, such as cybersecurity threats and issues of trust and control, emphasizing the need for research on safe performance. Technological unemployment is also examined, noting the potential for AI to displace jobs while arguing for optimism due to potential new job creation and the complementary nature of humans and AI.

1. Artificial Intelligence:
Progress and Challenges
Tom Dietterich

2. Tool AI
High-Stakes
Autonomy
Near-Term
Risks
?
Technological
Unemployment

3. What is AI?
Smart Software
 vision, speech, touch
 choosing actions to achieve goals
 learning
 understanding and predicting
behavior
Credit: Andrej Karpathy, Li Fei-Fei

4. Exciting Progress: Perception
2013 2014 2015
23% Word Error
8%
Google Speech Recognition
Credit: Fernando Pereira & Matthew
Firestone, Google
Google Translate from Images
Credit: www.bbc.com
“a black and white cat is
sitting on a chair."
Credit: Jeff Donahue, Trevor Darrell

5. Exciting Progress: Perception
Deep

6. Exciting Progress: Reasoning (SAT)
Credit: Vijay
Ganesh

7. Exciting Progress: Reasoning (Poker)
Moore’s
Law
Heads-Up Limit
Texas Hold’em
Rhode
Island
Hold’em
Credit: Michael
Bowling

8. Exciting Progress: Chess & Go
Monte Carlo Tree Search
Credit: Martin Mueller

9. Exciting Progress: Collaborative Systems
Red: Best AI solution
Yellow: “Foldit Void Crushers
Group” solution
Blue: X-Ray Crystal Structure
solution
Sept 20, 2011

10. What’s Next?
Tool AI
• Google, Bing
• IBM’s Watson
• Siri, Cortana, Google Now
• Google Translate
• Skype Real-Time Translation
Autonomous AI
• AI Hedge Funds
• High Speed Trading
• Self-Driving Cars
• Automated Surgical Assistants
• Smart Power Grid
• Autonomous Weapons

11. Tool AI

12. Tool AI: What’s Next?
Deeper understanding of video:
• What type of play?
• Who carried the ball?
• Was the pass complete?
• Which players made mistakes?
• Which players achieved their
goals?
Credit: Alan Fern

13. Tool AI: What’s Next?
Deeper understanding of text:
• Is Yoadimnadji dead? Yes
• Is he in Paris? Yes
• Is he married? Yes
• Where is his wife? In Paris
• Where will she be in the future? In
Chad
“Pascal Yoadimnadji has been
evacuated to France on
Wednesday after falling ill and
slipping into a coma in Chad.
His wife, who accompanied
him to Paris, will repatriate his
body to Chad.”

14. Tool AI: What’s Next?
Linking Big Data to Medicine:
• Web search logs can detect adverse
drug interaction events better than
FDA’s existing Adverse Drug
Interaction Reporting Service
White, Harpaz, Shah, DuMouchel, Horvitz, 2014

15. Tool AI: What’s Next?
Improved Personal Assistants that
combine
• Knowledge of recipes (lasagna
ingredients)
• Wine pairing recommendations
(cabernet or pinot noir)
• Brother’s home address
• Routes to brother’s home address
• Stores along that route that have
cheap cabernet or pinot in stock
to produce a plan
Source: Wired August 12, 2014

16. High-Stakes
Autonomy
Near-Term
Risks

17. Autonomous AI Systems: What’s Next?
• AI Hedge Funds
• High Speed Trading
• Self-Driving Cars
• Automated Surgical Assistants
• Smart Power Grid
• Autonomous Weapons

18. The Motivations The Dangers
High-Stakes Autonomy
 Advances in AI enable exciting
applications
 There is a great potential to save
lives
 There is a need to act at lightning
speed
 Bugs
 Cyber Attacks
 Mixed Autonomy
 Misunderstanding User
Commands

19. Software Quality
 Many AI methods give only probabilistic
guarantees
 Research Question: How can we ensure
safe performance of AI-based
autonomous systems?
"a young boy is holding a
baseball bat."
Credit: Andrej Karpathy, Li Fei-Fei

20. Cyber Attacks
 High-stakes systems  large and enticing “attack surface”

21. Cyber Attacks on AI Systems
Training Set Poisoning:
 Make yourself invisible to a computer
vision system
 Make yourself look “normal” to an
anomaly detection system
 Bias the behavior of the system
 Bid slightly higher on certain stocks
 Prefer to show certain advertisements
Credit: Katherine Hannah

22. Mixed Autonomy
 Auto-pilot unexpected hand-off to
pilots
 Pilots lack situational awareness
and make poor decisions
 Question: How can we make
imperfect autonomous systems
safe?
AF447 Pilots Final Conversation
Pilot 1: “What the...how is it we
are going down like this?”
Pilot 2: “See what you can do with
the commands up there, the
primaries and so on...Climb climb,
climb, climb
Credit: www.aviationlawmonitor.com

23. Incomplete/incorrect Commands
 “Fetch me some water”
 Question: How can the
computer reliably infer
what the user
intended?
Credit: www.disney.com

24. Trustable AI for Autonomy
Many research teams are at work
Verification and Self-Monitoring
Knowledge of Desirable and Acceptable Behavior
Improved User Interaction

25. Technological
Unemployment

26. 1980 1990 2000 2010
40
80
120
160
200
240
280
320
US Industrial Production
US Manufacturing Employment
Index:01.1972=100
Similar trends are seen worldwide
Source: Andrew McAfee / US Federal Reserve
US is Producing More with Fewer
Employees

27. As AI advances, many existing jobs have
the potential to be automated
Source: Frey & Osborne, 2014: “The future of employment: How susceptible are jobs to computerisation?”

28. Three Reasons to be Optimistic
 AI often complements human intelligence rather than
replacing it
 Increases in productivity increases in wealth
increases in consumption demand for new
goods and experiences  new kinds of jobs
 A focus on current jobs, underestimates the creativity
of people to develop new products and services, new
industries, etc.

29. What will be the new jobs?
 Jobs involving hard-to-automate skills
 high levels of social skills
 deep understanding of human experience
and emotion
 high levels of creativity
 Jobs where human + machine is better
than either alone
 Augmented Cognition
 combining strategic thinking with detailed
tactics: “Centaur Chess”
 combining physical dexterity with
information access: augmented reality
vehicle maintenance
Source: Tartajubow.blogspot.com
Source: metaio / designboom.com

30. Historical and cultural artifacts become
more valuable
 AI cannot produce another Roman
Empire, Greek Civilization, or Al Andaluz
 AI + Augmented Reality may make
tourism even more compelling and
meaningful
Credit: http://www.hellovisitspain.com

31. Life-Long Personal AI Assistant?
 Human-machine pairs work best when they
know each other very well
 strengths and weaknesses
 easy communication
 preferred ways of sharing tasks
 A vision:
 Student enrolls in technical college
 Student is given an AI personal assistant
 The student and the assistant train together
 Learn how to solve problems jointly and efficiently
 Employer hires the pair together
source: starwars.wikia.com

32. Tool AI
High-Stakes
Autonomy
Near-Term
Risks
?
Technological
Unemployment

33. Some people are very afraid
December 2, 2014
October 27, 2014

34. AI Misconceptions
Intelligence is Not a Threshold Phenomenon
 Progress in AI is the accumulation of thousands of incremental
improvements
 Robots will not “wake up” one day and be “truly intelligent” or
“superintelligent” or “conscious” or “sentient”
 “Tool AI” systems are already smarter than people along many
dimensions

35. AI Misconceptions
Autonomy Will Not Happen Spontaneously
 There is no threshold above which AI systems suddenly have free will
 Systems need to be designed and built to be autonomous
 They must be given access to resources (money, power, materials,
generalized task markets, communications with people)

36. The danger of
“Autonomous AI” is not “AI”
but “Autonomy”
An autonomous system can be dangerous for many reasons:
 It could consume vast resources
 It could injure or kill people
 It could apply AI to help it do these things

37. How Can We Maintain Control of
Autonomous Systems?
Case 1: Decisions are made at human time scales
Examples:
 Aircraft autopilot
 Most factory robots
 Surgical robots
Humans can monitor and respond to problems
 Similar to supervision in a human organization

38. Case 2: Very High Speed Decision Making
Examples:
 stock market
 power grid
 self-driving cars
 ≥18,520 market crashes and spikes
have been detected 2006-2011
below the 950ms level
Flash Crash (10.5.2010)

39. Automated Monitoring
 Signals might predict or
detect flash crashes
 Example: Inventory of High-
Speed Traders
Normal market Flash crash
Source: Vuorenmaa, Tommi; Wang, Liang, 2013

40. When Monitoring Detects A Problem:
Then What?
Stock market:
 Halt trading
 Unwind transactions
Self-driving car:
 Slow down and pull over in a safe spot
Smart Electric Grid:
 ??

41. We Should Never Create a
Fully Independent Autonomous System
By definition, a fully independent
autonomous system is a system over
which we have no control

42. Summary
 AI has been making steady progress
 Exciting applications of Tool AI are coming soon
 Potential applications of Autonomous AI are being proposed
 Many of these involve high-stakes decision making
 There are many risks that must be addressed before it will be safe to field
such systems
 Information technology—including AI—may be contributing to
unemployment but the picture is far from clear
 Tool AI will not spontaneously become autonomous
 We must find ways to control autonomous AI systems