Week 11 Study Guide: Environment and Moral Philosophy

Big idea

1. Data is material. The "cloud" and "clean tech" are stories. Behind them sit land, water, minerals, energy, and labor, much of it far away and made invisible.
2. The costs of tech are pushed onto others as externalities. Silicon Valley did this with chemical waste in the 1980s and is doing it again with data centers.
3. Environmental justice starts by undoing invisibility and externalization, then asks who holds power and what you can do.
4. Moral philosophy offers three Western frameworks: virtue ethics, deontology, and consequentialism. Each reveals different commitments and has its own blind spots.
5. Utilitarianism fits the datafied world best because both want to quantify, compare, and optimize from an impartial point of view of the universe.
6. Utilitarianism's classic failures, its logic of sacrifice, thin view of agency, and neglect of integrity and justice, show the limits of moral philosophy as a decision technology.

Sun, "No Data Centers in My Backyard"

Core idea: A road trip through Wisconsin and Michigan to understand why data centers became the most bipartisan hated issue in America. The answer is less about AI than about trust, process, and dignity. AI populism is mostly populism.

• Data centers are the most visible physical form of the AI boom: hundreds of acres, huge electricity demand, new transmission lines. About 70% of both parties oppose them.
• Janesville. A developer offered to clean up GM's toxic brownfield, but residents learned the talks had run a year under NDA. They forced a referendum and preferred an empty concrete slab to another single big employer.
• Mount Pleasant. Foxconn took over $4 billion in incentives, promised 13,000 jobs, and delivered about a thousand. Microsoft's data center on the same land was negotiated carefully, yet Foxconn taught residents never to trust a tech company or their own officials again.
• Port Washington. Residents said they were blindsided and passed a referendum requiring voter approval. The pro-data-center mayor blamed outsiders and would change nothing.
• Saline Township. Trustees voted no, the developer sued, and a town of 3,000 settled for $14 million. The utility redacted power estimates for a 1.4 GW site that adds 25% to its demand.
• Who benefits. Trade unions are thriving. Terry the union leader remembers when noise was money, and he bargains so his workers stay on site when automation arrives.
• Why the backlash. Four causes: material impacts that are real but similar to other industry, process failure where fast tracks and NDAs feel like cronyism, weak benefits with delayed taxes and few permanent jobs, and low trust in corporations and government that links local complaints to a national story. It's not Skynet, it's the oligarchy.
• Thymos. People want to be recognized as having worth, so bigger checks read as bribes. A backroom deal is corrosive because it says the public's buy-in never mattered. New York's statewide moratorium is the state stepping in to set terms.

Srinivasan, "Stop the Robot Apocalypse"

Core idea: A review of MacAskill's Doing Good Better. Effective altruism applies utilitarian math to charity and careers. Its logic leads from bed nets to funding AI safety, and it leaves the structures that cause misery untouched.

• The EA method. Do the most good, measured in QALYs, a universal currency that makes blindness and AIDS comparable. Think marginally and counterfactually: a doctor in a poor country adds 100x more value, but a banker who donates may save more lives than either, so earn to give.
• Replaceability. MacAskill once argued harmful jobs are fine because someone else would do them. He has backed off, since more bankers really do mean more banking.
• Capitalism's language. The book talks like global capitalism: quantify everything, treat all goods as commensurable, trust efficiency and the convergence of self-interest and morality. It mops up symptoms of inequality in the vocabulary of the disease and never asks how power works.
• X-risk. With 10^52 possible future humans, a tiny reduction in extinction risk outweighs saving a billion people now. So EA ends up funding AI research instead of malaria nets, which explains its appeal to Thiel and Musk.
• Too demanding, or wrong demands. Singer's drowning child says every luxury is murder. EA softens this to a 10% tithe, which dodges the objection but loses the distinctive insight. The deeper problem is that utilitarianism asks us to give up attachments and loyalties that make us human.
• Your life matters. You console a grieving friend because they are your friend, not because you met your quota. MacAskill refusing to fund the fistula hospital he visited as arbitrary is the point of view of the universe trying to step outside the only place ethics happens, your own.
• Individualism. EA takes the world as given and theorizes only the autonomous donor making marginal choices. It leaves everything as it is, which suits those who like the status quo.

Lecture 27: Environmental Contexts and Consequences of Data

Core idea: An environment-centered history of Silicon Valley, then the material footprint of data today. Resources are inputs, costs become externalities, and the work of invisibility is active.

• Clean image. San Jose boasted a skyline without smokestacks. Stanford Research Park and growth boosterism got built into infrastructure, zoning, taxes, and land annexation.
• Toxic reality. One 8-inch wafer needed 27 pounds of chemicals, thousands of gallons of ultrapure water, and 9 pounds of hazardous waste. The clean room protected chips, not workers or groundwater.
• Fairchild contamination. Leaking solvent tanks poisoned wells in San Jose, prompting the slogan Fairchild or my child. The Valley has 23 Superfund sites, once 29.
• Why so toxic. Doping and cleaning use toxic chemicals, petrochemical solvents were cheap, and scaling was fast. Plants sat in communities with less social power, worker health went unstudied, and regulation came decades later.
• Externality = a cost passed to people who don't benefit. Invisibility = excluding people or effects from view, which takes active work.
• Amazon Echo. Crawford and Joler's Anatomy of an AI System traces one voice command out to lithium in Bolivia, cobalt in the Congo, shipping, data centers, and e-waste. Every convenience rests on a planetary network of extraction.
• The cloud today. Data Center Alley in Northern Virginia, gas turbines powering xAI's Colossus in Memphis, land, water, and emissions. Environmental justice means making these visible and then asking where to start.

Lecture 28: Moral Philosophy

Core idea: Ethics asks how to live. Moral philosophy is the abstract, systematic, institutionalized version of that question. Three Western frameworks disagree on what makes an action good.

• Two dimensions: agency, who acts, who is responsible, what you owe, and the good, what a good life and good world require. Both are shaped by sociotechnical contexts.
• Virtue ethics asks what a person of good character does. Aristotle's phronesis and eudaimonia, a thick view of human nature. Critiques: which virtues, who decides, cultural specificity, conformism.
• Deontology is ethics as duty, judged by rule and intention. Kant's categorical imperative: act only on maxims you could will as universal law, and treat people as ends, not means. Freedom is autonomy.
• Consequentialism judges actions by outcomes, agent neutral, ends justify means. Utilitarianism is the version that maximizes utility: Bentham's greatest happiness for the greatest number.
• Trolley problem and MIT's Moral Machine put the frameworks to work on automated vehicles.
• Utilitarianism's reach. Quantification and objectivity, rational choice, choice architecture, risk scoring, panopticism, EA and longtermism. Sidgwick's point of view of the universe mirrors the impartial ideal of mechanical objectivity.

Lecture 29: Moral Philosophy 2

Core idea: Utilitarianism was co-produced with decision-making in the datafied world and resonates deeply in Silicon Valley. The classic objections to it also expose the limits of moral philosophy itself.

• Co-production. Game theory, the prisoner's dilemma, and Harsanyi's preference utilitarianism turned ethics into decision theory. Add self-tracking, venture capital, and QALY-based philanthropy.
• Silicon Valley. Longtermism and existential risk, with Sam Bankman-Fried as the case of EA logic ending in fraud.
• Classic objections:
  • Repugnant conclusion: a huge population with barely worthwhile lives beats a small happy one.
  • Logic of sacrifice: ends justify means, so some are sacrificed for the aggregate.
  • Experience machines and utility monsters: pleasure alone is not the good, and the math can justify eugenics.
  • Commensurability: not all goods and evils convert to one scale.
  • Justice: Rawls argues utilitarianism ignores how good is distributed.
  • Moral luck: outcomes depend on chance, as with the Uber self-driving death.
  • Integrity: Williams argues it treats your own projects as just another input.
• Hard choices and tragedy. AI, drones, and war show cases where no option is clean. Moral philosophy as a decision technology abstracts to a view from above and misses this.
• Alternatives: virtue ethics, existential ethics, and care ethics, which start from relationships and situated agents rather than impartial calculation.

How it connects

• Sun and Lecture 27 tell the same story forty years apart: a clean industry externalizes costs onto communities with less power, hides them, and calls resistance irrational. Fairchild's wells and Saline's gas plants are the same move.
• Srinivasan is the reading version of Lectures 28 and 29. EA is utilitarianism applied, x-risk is longtermism, and her objections, attachments, integrity, individualism, are the classic ones.
• Both halves of the week share one lesson: the impartial view from above, whether the cloud or the point of view of the universe, erases the situated people who bear the costs. Thymos and care ethics are both answers from below.
• Ties back: quantification and mechanical objectivity, Crawford's invisible labor, and the Valley's founder mythology now funding its own moral philosophy.