Skip to content
The Chat Archive
gemini-2.5-pro-preview-05-06· Google LMArena Human Preference 140K

People always say that automation creates more jobs? is it true? also, if AI gets really…

  1. User

    People always say that automation creates more jobs? is it true? also, if AI gets really really good, do you tihnk it can "create more jobs?"

  2. Assistant

    That's a fantastic and hotly debated question! The statement "automation creates more jobs" is a simplification of a complex historical trend, and its applicability to advanced AI is even more uncertain.

    Let's break it down:

    1. Historically, has automation created more jobs?

    Largely, yes, in the long run, but with significant disruption.

    • The Mechanism:

      • Productivity Gains: Automation makes producing goods and services cheaper and more efficient.
      • Lower Prices/Higher Quality: This can lead to lower prices for consumers or better products/services.
      • Increased Demand: Lower prices or better offerings can increase demand for those goods/services, requiring more production (even if each unit requires less labor, the total volume might increase).
      • New Industries & Roles: Automation often frees up capital and human labor to move into new industries and create entirely new types of jobs that didn't exist before. For example, the agricultural revolution displaced farm workers, but they eventually moved into factories. The rise of computers displaced typists but created jobs for programmers, IT support, data analysts, etc.
      • Supporting Roles: New jobs emerge in designing, building, maintaining, selling, and managing the automation itself.
    • The Catch (The "Disruption"):

      • Job Displacement: Specific jobs are definitely eliminated. Weavers lost jobs to power looms; switchboard operators to automated exchanges.
      • Skill Mismatch: The new jobs created often require different skills than the old ones. This can lead to periods of unemployment and hardship for those whose skills become obsolete.
      • Geographic Shifts: Jobs might move from one region to another.
      • Wage Stagnation/Inequality: Sometimes, the benefits of automation accrue more to capital owners than to labor, potentially leading to wage stagnation for some or increased inequality.

    So, historically, while painful for those displaced, the overall economic pie has grown, and new (often better-paying and less physically demanding) jobs have emerged. This is sometimes referred to as the "Luddite Fallacy" (the Luddites in the 19th century destroyed machinery fearing it would destroy their livelihoods, but overall employment grew).

    2. If AI gets "really really good," can it "create more jobs?"

    This is where things get much more speculative and potentially different from past technological shifts.

    • Arguments for AI creating more jobs (similar to historical patterns, but amplified):

      • New Industries & Frontiers: AI could unlock entirely new industries and scientific discoveries we can't even conceive of yet. Think personalized medicine at an unprecedented scale, climate change solutions, space exploration, AI ethics and governance, AI system trainers, AI content curators.
      • Human-AI Collaboration: Many new jobs could involve working with AI – guiding it, refining its outputs, providing human oversight, handling exceptions, and applying AI insights to complex problems.
      • Increased Productivity & Wealth: If AI dramatically boosts productivity, it could create immense wealth, leading to increased demand for all sorts of goods and services, including those that require a human touch (e.g., arts, crafts, personalized care, experiences).
      • Solving Unsolvable Problems: AI might help solve problems that currently limit human progress or well-being, freeing up resources and creating new opportunities.
    • Arguments why "really really good" AI might be different (and potentially lead to net job loss):

      • Cognitive Automation: Past automation largely replaced manual labor. AI is increasingly capable of automating cognitive tasks – tasks that were previously the exclusive domain of human intelligence (writing, coding, analysis, decision-making, even some forms of creativity). This is a much broader scope of potential displacement.
      • Speed and Scale: AI development is happening very rapidly. Society and education systems might not be able to adapt quickly enough to reskill the workforce.
      • Generality: If AI approaches Artificial General Intelligence (AGI) – meaning it can perform virtually any intellectual task a human can – then it's harder to see what "new jobs" humans would uniquely do, especially if AI can learn and adapt faster and cheaper.
      • "Jobless Growth": It's possible that AI could lead to massive productivity gains without a corresponding increase in the need for human labor, with benefits accruing primarily to those who own the AI.

    Conclusion:

    • Historically, yes, automation has, on balance and in the long run, led to more (and different) jobs. This involved significant, often painful, transitions.
    • With "really really good" AI, the answer is highly uncertain.
      • It could follow the historical pattern, creating new industries and roles we can't yet imagine.
      • However, because AI has the potential to automate cognitive tasks at scale, there's a real possibility that this wave of automation could be fundamentally different, potentially leading to a net reduction in the need for human labor, or at least a very challenging transition period.

    The outcome will likely depend on many factors, including:

    • The pace of AI development.
    • How we choose to integrate AI into society.
    • Policy decisions regarding education, social safety nets (like Universal Basic Income), and the distribution of AI-generated wealth.
    • Our ability to foster uniquely human skills like creativity, critical thinking, emotional intelligence, and complex interpersonal interaction.

    It's a future we need to actively shape, not just passively observe.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.