The Ethics of Artificial Intelligence - Chapter 10: AI Risks and Opportunities

Vocabulary flashcards covering the core concepts, theories, and philosophical arguments presented in Chapter 10 of Sven R. Nyholm's 'The Ethics of Artificial Intelligence', focusing on risk ethics, sustainability, and humanity's technological wager.

The AI Wager

The societal gamble of betting on the hope that artificial intelligence will solve major problems and lead to great prosperity, while accepting that it causes actual harms and brings significant risks.

Ethics of Risk

A philosophical subfield that examines when it is ethically permissible or impermissible to impose risks on other people, and what makes risk imposition morally problematic.

Risk (in Risk Ethics)

A potential negative side effect of an action characterized by two key elements: the probability that something bad will occur and the magnitude of harm or degree of badness associated with that outcome.

Decisions Under Uncertainty

Decision scenarios in which agents do not know how likely a negative outcome is or are unable to assess the magnitude of harm it would cause.

Pure Risk

Judith Jarvis Thomson's concept describing a situation where an agent's actions expose others to a risk of harm, but luckily the risk does not materialize and no actual harm occurs.

Russian Roulette Thought Experiment

A scenario where an individual pulls the trigger of a revolver with one loaded chamber at a sleeping person's head without harming them, used to show that exposing someone to pure risk is intrinsically wrong.

Sven Ove Hansson's Criterion for Risk Acceptance

The ethical view that imposing risks on others can be permissible if those risks are part of a shared practice that offers potential advantages and benefits to all or most members of society.

Longtermism

The philosophical perspective defended by Hilary Greaves and Will MacAskill asserting that moral agents should focus on the very far future because future generations possess equal moral significance to present people.

AI for Sustainability

Aimee van Wynsberghe's term for the practice of deploying artificial intelligence technologies to help solve environmental problems and achieve sustainability benchmarks, such as the UN Sustainable Development Goals.

Sustainability of AI

Aimee van Wynsberghe's term addressing the direct environmental footprint of AI itself, including substantial energy use, carbon emissions, cooling water consumption, and depletion of rare-earth minerals.

Moral Agents

Individuals or organized entities capable of performing actions that can be morally praised or blamed, and who bear a duty of care regarding the risks they impose on others.

Moral Patients

Beings who are on the receiving end of actions, whose interests matter morally, and who can be treated rightly or wrongly by moral agents.

Non-Identity Problem

Derek Parfit's philosophical dilemma noting that current large-scale societal choices determine which specific individuals will exist in the future, preventing future people from claiming those choices harmed them if they owe their very existence to them.

Drunk Driver Analogy

A philosophical analogy in AI ethics that compares rapid, unregulated AI development lacking safety protocols to reckless drunk driving, contrasting it with responsible, sober driving.

Utopian Optimism

The extreme perspective regarding the AI wager that assumes AI advancements will almost certainly produce immense abundance and widely shared prosperity without catastrophic side effects.

Dystopian Pessimism

The extreme perspective regarding the AI wager that assumes AI advances will most likely culminate in worst-case disasters, profound human misery, or the destruction of humanity.

Responsible AI Optimism

An approach that aims for the beneficial outcomes of AI while proceeding cautiously, implementing safety mitigations, and setting strict limits on permissible risk exposure.

Reckless AI Optimism

An approach that rushes AI development forward at maximum speed hoping for positive outcomes, while neglecting safety precautions or actively accepting severe risks as a means to achieve its goals.

AI Welfare

The emerging research field exploring under what conditions artificial intelligence technologies could become sentient or conscious, and what moral duties or rights humans might owe to them.

LLM Sentience

A concept defined by Google researchers Geoff Keeling, Winnie Street, and Blaise Agüera y Arcas as the capacity for valenced experiential states within large language models.