AI summary by TwinMind
The Rise of the Machines – Why Automation is Different this Time
Kurzgesagt – In a Nutshell
10m 57s
Historical Context of Automation
- 2013 study: almost half of all US jobs could potentially be automated in next two decades
- Past automation pattern: innovation → higher productivity → fewer old jobs + many new/better jobs
- Clear progression of human work: agriculture → industrial production → service jobs → information age
- Traditional automation worked well for majority, living standards improved
Current Automation Differences
- Jobs being taken over by machines much faster than past
- Information age industries creating fewer new jobs despite booming
- 1979: General Motors employed 800,000+ workers, made ~$11 billion
- 2012: Google made ~$14 billion, employed 58,000 people
- Car industry example: created huge industries 100 years ago, now innovation doesn't create as many jobs
- Electric cars won't create millions of new jobs
Internet Innovation Impact
- Internet compared to electricity introduction in transformative power
- Creates new industries but not enough jobs for population growth
- Peak 2004: Blockbuster had 84,000 employees, $6 billion revenue
- 2016: Netflix had 4,500 employees, $9 billion revenue
- Kurzgesagt example: 12-person team reaches millions vs TV station needing way more employees
- Information age innovation ≠ creation of enough new jobs
Machine Learning and Digital Automation
- Human progress based on division of labor, jobs became specialized
- Machines excel at narrowly defined, predictable tasks
- Complex jobs = many narrowly defined predictable tasks in sequence
- Digital machines use machine learning: acquire information/skills by analyzing data
- Humans gathering data about everything: behavior, weather, medical records, communication, travel, work
- Created huge library for machines to learn human tasks
- Digital machines can be replicated instantly and for free
- Computer workers might lose jobs sooner than factory workers
Real-World Example: San Francisco Project Management Software
- Eliminates middle management positions
- Process: decides which jobs can be automated → assembles freelancer teams → distributes tasks → tracks performance
- Learning algorithms track freelancers, gather data about their work
- Freelancers teaching machine to replace them
- Reduces costs ~50% first year, additional 25% second year
- Similar patterns across fields: pharmacists, analysts, journalists, radiologists, bank tellers, burger flippers
Job Creation Crisis
- Need constant new job generation due to growing world population
- Since 1973: new job generation in US began shrinking
- First decade 21st century: first time total US jobs didn't grow
- US needs 150,000 new jobs/month to keep up with population growth
- Affecting living standards
Productivity vs Employment Data
- 1998: US workers worked 194 billion hours total
- Next 15 years: output increased 42%
- 2013: still 194 billion hours worked (no growth)
- Despite: productivity growth, thousands new businesses, 40+ million population increase
- New university graduate wages declining past decade
- Up to 40% new graduates forced into jobs not requiring degrees
- Productivity separating from human labor
Economic Implications
- Automation different this time - machines might really take jobs
- Economies based on premise people consume
- Fewer people with decent work = who will consume?
- Risk: producing cheaply but too few people can buy products/services
- Potential future: tiny minority super rich owning machines dominating rest
Future Outlook
- Information age/modern automation could be huge opportunity
- Potential to change human society, reduce poverty/inequality drastically
- Could be seminal moment in human history
- Part two will discuss potential solutions including universal basic income
- Machines not coming - already here
