This is a genuinely contested question among economists, and the honest answer is: it depends on timeframe, industry, and policy response—and experts disagree. Let me lay out the main perspectives rather than claim a settled answer.
The case for "takes away more"
•Routine/predictable tasks are most exposed. Jobs involving repetitive data processing, basic customer service, transcription, routine coding, and standardized writing are already seeing automation. Generative AI specifically threatens white-collar "knowledge work" that previous automation waves spared.
•Speed of disruption matters. Even if AI eventually creates jobs, the transition could be faster than workers can retrain, causing real hardship for specific occupations (e.g., some customer support, paralegal, copywriting, and entry-level coding roles are already shrinking in some companies).
•Concentration of gains. Productivity gains may accrue mostly to capital owners and a smaller set of highly skilled workers, rather than being broadly redistributed into new job categories for displaced workers.
The case for "creates more" (or roughly offsets)
•Historical pattern: every prior automation wave (agriculture mechanization, computers, the internet) destroyed specific jobs but created more new ones than it eliminated—often in categories unimaginable beforehand (web design, app development, social media management, data science).
•New job categories already emerging: prompt engineering, AI ethics/safety roles, AI auditing, human-AI workflow design, and massive growth in AI infrastructure (chips, data centers, energy).
•Productivity → growth → demand: If AI makes companies more productive and profitable, that can fund expansion, new products, and new hiring elsewhere, especially if cheaper goods/services increase overall demand.
•Human complementarity: Many roles may shift rather than disappear—AI handles routine subtasks while humans focus on judgment, creativity, relationship-building, and oversight.
Key uncertainties that will decide the outcome
1.How fast AI capabilities advance relative to how fast new industries/jobs can form.
2.Policy choices: retraining programs, education systems, labor regulations, and possibly how gains are taxed/redistributed.
3.Which "ceiling" AI hits: if AI remains a powerful tool requiring human oversight, complementary jobs likely grow; if it becomes broadly autonomous across cognitive tasks, displacement could dominate for longer.
4.Global variation: effects will differ drastically by country, industry, and skill level—manual/physical jobs, highly social jobs, and highly regulated/trust-dependent jobs (nursing, skilled trades, therapy) are more insulated than routine cognitive work.
Bottom line: There's no consensus answer—serious economists, labor researchers, and AI researchers land on different sides depending on their assumptions about the pace of AI progress and the ability of labor markets and policy to adapt. If you want, I can pull recent data/reports (e.g., from the IMF, WEF, McKinsey, or labor economists) on which specific sectors are seeing net job growth vs. decline right now, which might be more useful than the overall theoretical debate.