Loading book details
Loading book details
A World Without Work cover

A World Without Work

Daniel Susskind•2020

  1. Chappy's Book Notes•332 books

A World Without Work

Daniel Susskind•2020

Length
9h 56m•~328 pages
Read
May 3rd - 4th '23
AIFuturismPoliticsMacroeconomics
•

Summary

Automation is coming for tasks > jobs and the distinction matters. AI makes tacit knowledge explicit through pattern recognition, steadily encroaching on manual, cognitive, and affective capabilities that were once considered uniquely human. The result won't be mass unemployment overnight but instead a combination of frictional unemployment and eventual structural unemployment where there simply aren't enough tasks left for humans to justify wages. The response requires not just education reform but a fundamental rethinking of distribution – conditional basic income, better taxation of capital and data, and new sources of meaning beyond work.

Key Takeaways

  • Automation is task-based, not job-based – the ALM hypothesis
  • AI's pragmatist revolution shifted from copying human cognition to optimizing task performance
  • Systems are opaque because their patterns don't need to make sense to us
  • Decoupling of human/traditional capital
  • Mismatch of skills: workers can't do the new jobs
  • Mismatch of identity: workers won't accept lower-skill roles
  • Mismatch of place: jobs exist but in the wrong geography
  • Complementing effects (productivity, bigger pie, changing pie) have diminishing returns as AI capabilities grow
  • Humans may become like horses after 1900 – capabilities simply overshadowed
  • The great decoupling: 1995-2015 saw 30% productivity growth but only 16% pay growth
  • Need to tax workers benefiting from automation, capital, robots, inheritance, and companies
  • College return is ~15% but 80% of that value is signaling, not skills
  • Conditional basic income with contribution obligations (elder care, arts, education) preserves purpose
  • Hunter-gatherers had 5-6 hours of daily leisure – work-as-identity is historically unusual
  • Roots of maladaptive behavior often trace to childhood trauma and perceived helplessness
  • Purpose must be actively constructed through community obligations, education, and creative contribution

Notes

1: The context

1: A history of misplaced anxiety

  • Total wealth → inequality, power, purpose
  • Substituting vs complementing effects

2: The age of labor

  • Skill premium: HS vs college education pay
  • ↑ historically hasn’t always favored high-skill jobs like it does now
  • Explicit vs tacit knowledge jobs
  • Moravec’s paradox
  • Automation is task-based, not job-based
  • ALM hypothesis

3: The pragmatist revolution

  • Early AI: just copy humans
  • Pragmatist revolution: AI based on task performance, not means of solving
  • Shift away from using AI research to understand human cognition
  • AI researchers ↔︎ religious theologians
  • Modern AI research akin to evolution: “blind watchmaker”, not divine creation

4: Underestimating machines

  • AI makes tacit knowledge explicit (patterns)
  • Systems are opaque because patters don’t make sense to us (and shouldn’t have to)

2: The threat

5: Task encroachment

  • Good task: clear objective, lots of data
  • 3 main human capabilities:
    • Manual: physical world
    • Cognitive:
    • Effective: interpersonal
  • AI will keep encroaching on new tasks

6: Frictional technological unemployment

  • FTE: jobs are there, but can’t taken up
  • Three reasons:
    • Mismatch of skills
    • Mismatch of identity: to lower skill
    • Mismatch of place: geography
  • Job market has become polarized: middle skill jobs are getting hollowed out

7: Structural technological unemployment

  • STE: not enough jobs to do
  • Three ways complimenting affect raises number of jobs:
    • Productivity effect: only relevant when still cheaper than automation, human has something to contribute
    • Bigger pie effect: not necessarily rising demand for humans
    • Changing pie effect: ↑
  • Humans like horses of 1900s as capabilities overshadowed by AI
    • Horsepower → manpower

8: Technology and inequality

  • Traditional vs human capital
  • Gini coefficient: rising, rich getting richer
  • Historic economy by capital source:
    • Traditional: 1/3, human labor: 2/3
    • “One of 6 key Econ facts”
    • Ratio is breaking since 1980s
  • The great decoupling: ↑
    • ’95-’15: ↑ 30% productivity, 16% pay
  • % rise in inequality due to technology:
    • OECD: 80%, IMF: 50% (25% global)
  • Countries can shape inequality

3: The response

9: Education and its limits

  • Average college return rate: 15%
  • Need to teach the right topics
  • Personalized learning
  • College value: 80% signaling, 20% skills

10: The big state

  • Free markets vs central planning
  • Big state: not production, but distribution
  • Need to tax workers benefiting, capital (monetary, robots), inheritance, companies
  • Better laws, accounting code of conduct
  • UBI: citizens income; what level?
    • Conditional basic income (CBI)
    • Contribution obligation: eg. elder care

11: Big tech

  • Need data, software, and hardware
  • Network effects
  • Monopolies entrenching themselves
  • Big tech outsized influence on society
  • Self-regulation issue

12: Meaning and purpose

  • Work: purpose, meaning, status, esteem
  • ↑ not always the case (hunter gatherers)
  • 5-6 hours a day of leisure
  • CBI: actives society requires them to do
    • Education, social care, arts
  • Roots of behavior are childhood trauma
  • Trauma: perceived helplessness
  • 5 types of trauma:
    • Abuse: physical, sexual, spiritual
    • Neglect
    • Abandonment
    • Enmeshment: blurring child/adult life
    • Witnessing tragic events
  • Trauma vs adversity
  • 4 branches of maladaptation:
    • Addiction: vices, work, perfectionism
    • Codependence: excessive dependency
    • Habituated survival strategies
    • Attachment disorders
  • “90% of male rage is helplessness masquerading as frustration”
  • Reframing: self-centeredness
  • Therapy, nature