Bingecast: Is Cheese Consumption Causing Deaths from Tangled Sheets?
Those dealing with data must always remember “If you torture data long enough, it will confess to anything.” The answers that computers give must themselves be questioned. Robert J. Marks and Gary Smith address artificial intelligence, spurious correlations, and data research on Mind Matters.
Show Notes
- 01:34 | Introduction to Gary Smith, the Fletcher Jones Professor of Economics at Pomona College
- 02:40 | The AI Delusion
- 05:04 | Stocks, Data, and Analysis
- 07:04 | Presidential campaigns and data analytics
- 10:53 | Bill Clinton and neural networks
- 11:28 | IBM Watson and Jeopardy
- 13:30 | The Winograd Schema Challenge
- 15:06 | Bob Dylan and IBM Watson
- 17:00 | Understanding words
- 19:00 | John Searle’s “Chinese Room” and understanding numbers
- 20:08 | Following instructions versus understanding. “Is it safe to walk downstairs backwards with your eyes closed?”
- 21:34 | IBM Watson’s ineffectiveness in health and medicine
- 24:15 | What is the crux of Watson’s failure?
- 25:31 | Chamath Palihapitiya’s brutal verdict on Watson
- 26:28 | “Artificial Intelligence,” the 2017 marketing word of the year
- 26:57 | The “algorithm of the gaps”
- 27:39 | Roger Schank and Douglas Hofstadter seeking human thought
- 28:07 | Solving narrow problems to make money
- 28:34 | “Climbing a tree to reach the moon”
- 28:43 | General intelligence, i.e. common sense, the Holy Grail of AI
- 30:14 | Bob’s thesis: computers can only ever do algorithmic things
- 31:11 | Skepticism of Edward Leamer, statistician at UCLA
- 32:03 | Spurious correlations
- 33:34 | The Smith Test
- 35:35 | AI is a black box
- 37:56 | Differentiation between wolves and dogs
- 38:29 | Fooling face recognition
- 41:36 | Fooling humans with random data
- 42:41 | The bigger the data, the more spurious correlations
- 45:05 | Data mining and the origin of the term “Texas Sharpshooter Fallacies”
- 46:05 | Fallacy #1: Multiplying targets
- 47:06 | Drinking coffee and pancreatic cancer
- 48:30 | The profusion of health studies and claims
- 50:08 | Fallacy #2: Drawing the target after shooting
- 51:27 | The pressure in academia to publish or perish
- 51:43 | A story from J. B. Rhine’s ESP lab
- 52:59 | Diederik Stapel and fabricated data
- 53:22 | Spurious correlations in big data
- 54:14 | John Ioannidis, the “decline effect,” and the status of flawed medical research
- 55:29 | The health media’s headline clickbait
Additional Resources
- Gary Smith at Pomona College
- Standard Deviations: Flawed Assumptions, Tortured Data, and Other W by Gary Smith
- Money Machine: The Surprisingly Simple Power of Value Investing by Gary Smith
- The AI Delusion by Gary Smith
- Claude Shannon at Encyclopædia Britannica
- The Winograd Schema Challenge
- Roger Schank on IBM Watson
- Bob Dylan and IBM Watson on the meaning of his lyrics
- On John Searles’ “Chinese Room Argument” at the Stanford Encyclopedia of Philosophy
- The Association of National Advertisers announce the “Marketing Word of the Year For 2017“
- Roger Schank at Socratic Arts, Douglas Hofstadter at Indiana University, Oren Etzioni at the Allen Institute for Artificial Intelligence, Edward Leamer at UCLA
- Spurious correlations website
- “The Mind of a Con Man” at The New York Times on Diederik Stapel’s academic fraud
- John Ioannidis and the “decline effect” at Wikipedia