AI's Dark Side: Hype Bubbles, Job Loss, and Control? (lightcapai.medium.com)

AI use has already scaled from novelty to infrastructure: an analysis finds ChatGPT alone handles roughly 18 billion messages per week, with weekly query volumes like ~1.9B for writing/editing, ~1.8B for tutoring, ~1.0B for health/self-care, ~0.75B for coding help and ~0.7B for creative ideation—orders of magnitude beyond what human professionals provide. That abundance creates enormous opportunity (ubiquitous tutoring, private health guidance, faster development workflows) but also acute risks: many “how-to” prompts hide high-stakes questions, and the raw volume means AI is already part of intimate decision-making at scale. The piece warns of two emergent commercial paths with technical implications: premium scarcity (gating top models behind paywalls and compute quotas) and “free” with hidden monetization (using private prompts to train models, target ads, or bias outputs toward partners). Both produce inequality, privacy loss, and incentive-driven output bias. A parallel threat is the content assembly line—AI-powered editing and text-to-video tools (e.g., YouTube features) that optimize for short-term engagement metrics. Early data show low completion rates (~11%) but high tolerance for AI-generated clips (~63% willingness), signaling a feedback loop where algorithmic best-practices produce homogeneous, high-volume “AI slop,” displacing creators and eroding diversity. For the AI/ML community this underscores urgent priorities: governance and data-provenance standards, transparent monetization and evaluation metrics beyond click-throughs, and research into equitable access and robustness to incentive-driven bias.