Long-read argues AI companies should fund their externalized costs now, citing Summitville, tobacco and nuclear trust models
Hatter_of_Time · reddit · 2026-11-03
A Reddit long-read uses historical precedents to argue AI companies should fund externalized costs now, because "the cheapest day to pay is today":
- Historical pattern: Summitville gold mine (1986) posted <$5M reclamation bond, went bankrupt in six years, leaving a $240M public cleanup bill; tobacco delayed for decades then committed $206B over 25 years
- Two resources being hollowed out: physical (data centers' water/power/land — Frederick County, MD rejected >$110M in community benefits and extended a pause) and cultural (Stanford reports 22-25yo employment in AI-exposed occupations 19% below trend, narrowing the expertise pipeline — correlation, not proven cause)
- Workable model: US nuclear plants must fund their own decommissioning via a fraction-of-a-cent per kWh fee into an independent trust
- Three-step proposal for AI: Count (disclose energy/water use and training data, as New Jersey already requires), Fund (use-based fee into an independent trust, heavy users pay most, small users exempt), Adjust (reassess as real costs emerge — Alaska's 1976 oil fund is now worth >$80B)
The accompanying chart compares who paid for cleanups across industries: nuclear 100% industry-funded, Superfund public share rose from 20% to 53% after industry taxes expired, Summitville 86% public.
More from AGI Musings
- Google Ships Top AI Model, Stock Craters Same Week; Humanoid Robot Shipments Up 400% YoY — PeterDiamandis · 2026-11-03
- Hot take: robotics value will accrue to users of robots, not robot makers — ZeYanjie · 2026-11-03
- Sovereign AI is creating the sharpest haves-vs-have-nots split in 30 years, says investor — arieljalali · 2026-11-03
- DHH: programmers may be worse at AI coding than non-coders — '100%' — 2C_ornot2C · 2026-11-03
- Hypothesis: high-progress places tend to have lots of cults, from Bauhaus to AI safety — Afinetheorem · 2026-11-03
- Researchers propose adding Readability to ARR review criteria amid LLM-generated paper flood — TuhinChakr · 2026-11-03