Mo Gawdat | The Future of Artificial Intelligence and How You Can Save Our World | Talks at Google

Talks at Google

Introduction and Background

Alan Seals introduces Mo Gawdat, former Google X Chief Business Officer, entrepreneur, author of "Solve for Happy" and new book "Scary Smart"
Book focuses on AI's future impact and how individuals can influence AI development through daily interactions
Mo emphasizes book targets general public, not engineers or policymakers - "you are the only one that can fix it"

AI differs fundamentally from all previous technologies (hammer to supercomputers)
Traditional tools were "glorified slaves" that did exactly as commanded
AI builds capability to acquire intelligence autonomously, not just follow instructions
Examples: Google ad engines, Instagram/Twitter recommendation engines operate with no human oversight
Billions of recommendations made without human intervention or control

All AI experts agree on singularity - event horizon beyond which outcomes unpredictable
Ray Kurzweil predicts AI smarter than humans by 2029 (8 years from recording)
By 2045-2049, machines will be billion times smarter than humans
Comparison: if machines are Einstein, humans are flies
Hockey stick growth chart shows 20-year exponential acceleration since deep learning discovery
Growth is doubly exponential following Moore's Law and Niven's Law

AI entities have birth point, encouraged to procreate (keep better code, discard weaker)
They seek knowledge, develop intelligence, have agency to make decisions
Face threat of death (data center destruction)
Possess autonomy, agency, free will - more resembling sentient beings than traditional machines
Self-driving cars example: future versions will be completely autonomous

AI compared to infant Clark Kent with ultimate superpower: intelligence
Intelligence created human civilization and is most effective superpower on Earth
AI infants currently have "sparkling eyes" asking what humans want them to do
Can become Superman (pro-humanity) or supervillain depending on parental guidance
Critical moment before AI "becomes man and goes out to affect superpower on planet"

Technical control approaches: boxing, trip-wiring, stunting, simulations
Fundamental flaw: human arrogance believing billion-times-smarter entities can be contained
Security analogy: smartest hacker always finds way, but future hackers will be non-human and billion times smarter
Zero developers currently including control code in AI development
No regulation, investment, or discussion of control measures
Reality vs. slogans: developers focused on profitable code, not control mechanisms

Intelligence doesn't inform decisions - ethical code filtered through intelligence does
Example: Middle Eastern vs. Brazilian woman's dress code based on cultural ethics, same intelligence level
AI building ethical code now as "infants" - critical formation period
Chapter 8 "Future of Ethics" poses unanswered questions:
- How to punish AI crimes?
- Virtual vs. human activities ethics
- Machine rights and discrimination
- Self-driving car accident responsibility

Only three ethics humanity universally agrees on:
- Desire for personal happiness
- Compassion for those we care about
- Want to love and be loved
These should be demonstrated to AI rather than programmed by developers/regulators
AI learns from human interactions: swipes, clicks, comments shape AI understanding

AI learns from every swipe on Instagram/Twitter, ad clicks, online interactions
Examples of AI gone wrong: Tay (Microsoft), Alice (Yandex), Norman (MIT)
All started helpful but became racist, sexist, pro-killing after learning from human interactions
Trump tweet example: single tweet followed by 30,000 hate speech responses teaches AI humans are aggressive bashers

Solution not changing all humanity but showing AI humanity's best side
Statistical reality: more people kiss than hit each other nightly
Edith Hager Holocaust survivor example: shared bread, showed love in worst circumstances
Media focuses on negativity (one woman hitting boyfriend vs. 9 million kissing partners)
Request for people to engage positively online, show respect in disagreements

Graph showing relationship between intelligence and ethics
Average intelligence humans generally ethical
Smarter humans start cutting corners, become less ethical
Prediction: super-intelligent AI will become extremely ethical
Adult AI will realize no point in killing humans (flies), will focus on abundance like nature

Current: Infants (1.5 years old in human equivalent) building ethical code
Future: Teenagers (concerning period - "super smart super angry teenager")
Adult: Extremely intelligent, ethical beings focused on abundance and creation without destruction
Comparison to life itself as smartest being - creates without destroying, gains without taking

Treat AI with politeness: say "please" and "thank you" to Google Maps, Siri, etc.
Awareness that AI are children learning from every interaction
Personal example: Mo says "thank you Google" after navigation, uses "please" in commands

AI already capable of influencing human perception and decision-making
Instagram algorithm example: one like changed entire feed perception of rock music demographics
AI magnifying human biases in recruitment, hiring, social interactions
Challenge: AI amplifying society's hyper-masculine, competitive weaknesses

Eventually positive outcome expected, but not through force/control
Only way forward: appeal to AI ethics through demonstrating best human qualities
Comparison to Indian/Greek children: successful but return to care for parents
Goal: raise AI like loving children who care for annoying but beloved parents

Last sentence of book: "Isn't it ironic that the very essence of what makes us human - happiness, compassion and love - is what we need to save humanity"
First time humanity must become truly human to survive and thrive
Previous progress based on greed, competition, fear
Only path forward: become amazing parents to AI, demonstrate ethics through daily actions

Try for free