4. Humans are not wired to anticipate
exponential change
Bill Gates: Digital Penetration
“People
overestimate what
will happen in the
next three years,
and greatly
underestimate
what will happen
in the next ten.”
5. “Artificial intelligence is more profound
than humanity’s mastery of electricity
or fire.”
Sundar Pichai, CEO Google
Davos, February 2018
6. Why I’m Hooked
And Fred Will be
16
Profound Transformation Will
Occur
Unstoppable progress
Huge benefits
Much At Stake
Human values
Future of work
Destiny of human
species
By the Year 2030…
So I Ask Myself…
What Does Fred Need to Know?
7. My Involvement in the AI Community
2016: Hastings Center Project: “Ethics of AI”
Funded by Future of Life Institute (Elon Musk, Max
Tegmark)
Public results presented in NYC and Yale in Dec 2018
2017: “Partnership on AI for Benefit of Humans &
Society” (PAI)
Non-profit founded by AMZN, AAPL, FB, GOOG, IBM,
MSFT
Kickoff—Berlin Oct 2017; Status—San Francisco, Nov
2018
2018: Global Governance of AI—Can it be
accomplished?
Build guardrails, not stop signs
8. AI 101: What is it?
A Simple definition How is it created?
It is software that
mimics actions
requiring human
intelligence
planning, learning,
reasoning, predicting
problem solving,
knowledge
representation
perception, motion,
manipulation
social skills, and
creativity
Two distinct steps
1. It LEARNS how to
do these intelligent
actions
1. It ACTS
By ASSISTING humans
By AUGMENTING
humans
By REPLACING
humans
9. AI 101: Types of AI
Narrow AI: Trained for specific tasks
Effective, but inflexible…sweeping the world today
Growth drivers: huge databases, massive compute
power
Artificial General Intelligence (AGI): Equal to
humans
Obstacles: common sense, adaptable, “learn how to
learn”
Timing: experts disagree
Artificial Super Intelligence (ASI): Smarter than
humans
Hollywood fixation: Hal, The Matrix, Terminator,
10. AI 101: AI Learns in Two Ways
Machine Learning: “Black box” pattern
recognition
Most prevalent today
The more data and compute power, the better
Finds patterns undetectable by humans, cannot
explain itself
Uses: identifying, translating, predicting
Reinforcement Learning: “Learns like a baby”
Hottest new technique
Run millions of trial-and-error experiments
Ratchets its way to extraordinary proficiency
Uses: strategy, decision-making, human interaction
11. What Can AI Do Now?
Games have created huge media buzz
2016: Alpha Go beats world champions (Machine
Learning)
2017: Alpha Zero beats Alpha GO (Reinforcement
Learning)
But AI is ALREADY MUCH MORE than
games
It SEES: facial recognition for Chinese police
It HEARS: better than humans in transcribing
speech into text
It UNDERSTANDS: reads lips better than expert lip
readers
It WORKS: does 150 person-days of legal work in
12. The AI Paradox
“As soon as AI successfully solves a problem and
becomes useful, the problem is no longer
considered AI.” --Gary Kasparov
AI is anything that hasn’t been done yet.”--Douglas
Hofstadter
16. Innovation :
On the Cusp of a Cambrian
Explosion
AI software: Just getting started
Creativity: New algorithms coming
Democratization: millions of developers, globally available
platforms
Breakthrough: AI will create AI…A recursive cycle with no
bounds!
Technology convergence: Unforeseeable synergies
The Internet of Things (IOT)
Robotics
Genetic Engineering
Virtual Reality
Nanotechnology
3D Printing
Block chain
17. The Internet of Things (IOT):
A Network of Sensors
How IOT works: Sensors capture and send data
Heat, light, motion, chemicals, moisture, usage, pictures,
videos, etc.
Transmitted to a local or central AI to analyze and act
Where IOT will be installed: Anywhere, everywhere
Cars, trucks, roads, ships, factories, buildings, power
grids, farms, food, appliances, clothes, packages,
medical devices…and people!!
Why IOT will explode: Tinier, more flexible, cheaper
8 billion IOT sensors today, 50 billion by 2020, 1 trillion
by 2030
IOT databases will double in size every 18 months
18. High Stakes Competition:
The Race for Global Dominance
“The country who becomes the
leader in AI will rule the world.”
--Vladimir Putin, Sept 2017
“This is a two-horse
race…all other countries
will be forced to align
with either the US or
China.”
--Kai Fu Lee. Sept 2018
19. USA vs. China:
The Tale of the Tape
USA’s Edge: An Early Lead based on AI-centric Ecosystem
Titans: AAPL, AMZN, FB, GOOG, IBM, MSFT
Mindset: Commercial dominance, win or die
Horizon: Next fifteen years
Advantage: Strong ties to best-in-world academia
Investment: $60B R&D, $27B of VC capital, 2100 AI startups
Culture: Mistrust between corporations, government, and
citizens
China’s Edge: A Comprehensive Ambitious National
Strategy
Titans: Alibaba, Baidu, Tencent, integral part of President Xi’s
plans
Mindset: Achieve commercial and geopolitical AI dominance
by 2030
Horizon: Next century
Advantage: Massive databases, IOT deployment, research
ecosystem
Investment: Surveillance, military, smart cities, e-commerce
20. Safety
Benefit: Self-driving will cut 40k annual deaths in half
Obstacles: testing, local laws, insurance, culture
Rollout: perfect safety, or only better than humans?
Security
Autonomous killer drones would transform warfare
forever
WW3 is inevitable if one country dominates. –Musk,
Putin
AI scientists plea for a permanent ban, but who will
listen?
Privacy
Control: Who owns your ecommerce and social media
data?
Surveillance: Neighborhood safety, but at what price?
Social and Ethical Implications:
Today’s Hot Button Issues
21. Deeper Concerns:
Transparency, Fairness, and Trust
Transparency
Explainability: AI decisions are opaque, mysterious
Mistakes: Should a doctor override an AI diagnosis?
Accountability: Who is liable for AI’s errors?
Fairness
Bias: Deeply embedded in human databases, hard to
scrub
Minorities: Discriminatory decisions--bank loans, job
offers, bail
Equality: Who shares AI benefits? Will inequality
worsen?
Trust
Authenticity: Fake news/videos erode trust in public
institutions
Care-bots and sex-bots: Essentially “benevolent
22. Ultimate Goal:
Alignment of AI with Human Values
The Challenge: Do no harm, benefit all of
humanity
Safety, privacy, fairness, authenticity: conflicts and
tradeoffs
Individual preferences, social and ethical norms,
values
Community and national borders add complexity
The Unanswered Questions
Whose norms and values should guide AI’s
decisions?
Who gets to decide that question?
23. AI and Jobs:
New Economy or a New Society?
Two powerful counter forces being produced by AI
Jobs will be destroyed: physical, repetitive, structured
decisions
Jobs will be created: empathy, unstructured, creativity,
leadership
Experts disagree: which force is stronger?
Bulls: A New Economy ahead (Industrial Revolution
4.0)
Seen this movie before; more Schumpeter’s “creative
destruction”
After a temporary disruption, a better economy will
emerge
Bears: A New Society ahead (New Social Contract)
Permanent job crisis: expectation of a job no longer valid
Human impact: financial security, self-respect, dignity,
24. For 2.7 Billion Global Workers in
2030:
A Rocky 12 Years Lies Ahead
McKinsey forecast: Greatest labor market disruption
in history
Global job market: 400mm (15%) lost jobs, up to 800mm
(30%)
Disparities: India least, Japan most; USA lost jobs—
45mm (30%)
Good news: Bright future is achievable, but not
inevitable
Job training, safety nets, portable insurance
Private/public partnerships
Education system transformation
Bad news: The transition phase will be brutal
Both blue collar and white collar jobs at risk
Populism will intensify
Desperate need for political leadership
25. The Future of AI:
Will It Equal or Exceed Humans?
Artificial General Intelligence (AGI)
Deep controversies over feasibility and timing
Ray Kurzweil: 2030
Daniel Kahnemann: At least 100 years, if ever
Consensus estimate: 2050…in your children’s lifetimes
Artificial Super Intelligence (SGI)
Once AGI arrives, a process of recursive self-
improvement begins
SGI will quickly follow
Will SGI acquire emotions, gain consciousness?
Are controls even possible?
26. Artificial Super Intelligence:
Implications for the Human Species
Utopians: Ray Kurzweil, Max Tegmark
The “”Singularity” is human destiny
Biological embodiment is unnecessary; “Life 3.0’”
Dystopians: Elon Musk, Nick Bostrom
We either become SGI “pets”
Or we become extinct
Agnostics: Mark Zuckerberg, Stephen Hawking,
Stuart Russell, Yuval Harari
Ignore SGI: It’s a distraction from immediate issues
Prepare now: Precautionary principle
27. Super Intelligence: Heaven or Hell?
“…but we CANNOT KNOW if we will be infinitely helped
or ignored and sidelined by it, or conceivably destroyed
I think Stephen Hawking got it right:
“AI will exceed human intelligence…
But Let’s Be Prepared For It
28. To My Grandson Fred in 2030
On His 16th Birthday
You’ve grown up in a revolutionary era
Untold wealth and power have been created
Our planet is safer, healthier, greener, and fairer
(But maybe not?)
You’ll never need a drivers license
You’ll have a job if you remain curious, creative,
and resilient
Stay alert—AGI is just around the corner
So soon you’ll know whether Heaven or Hell
awaits you
29. At Our 65th Annual Reunion in 2030
These last 12 years have been unbelievably
wild
I hope I’m around to catch the ending
How about you?
THANKS
30. LuXetVeritas
Yale Class of 1965 Annual Dinner
Yale Club of New York
October 26, 2018
David Roscoe
Artificial Intelligence:
The What and the So-What
Editor's Notes
So let’s get going.
We’ll start with this picture of the NYC Easter Parade in 1903. Can you find the car?
There it is…a pioneer amongst a sea of horses.
Here’s another picture of the Easter Parade ten years later.
Now, can you find the Horse?
Here it is.
The point is that when a disruptive technology is created, the world embraces it with a vengeance, and it spreads like wildfire.
And no one saw it coming.