Can we code ethics into machines? | Dila Şen | TEDxEsade Salon

Quick Overview

The presentation concludes that the ethical challenge in AI, exemplified by autonomous vehicles and lethal autonomous weapons, is not primarily technical but stems from humanity's historical tendency to outsource moral responsibility to authority figures or abstract systems, necessitating the development of explicit, universally agreed-upon ethical frameworks rather than relying on flawed human precedents.

Key Points: The core ethical dilemma for AI, illustrated by self-driving cars and autonomous weapons, is deciding who the machine should sacrifice in an unavoidable accident (e.g., one person vs. five people, or a child vs. a lobster). Historical human precedents, such as the Milgram obedience experiments (1960s) and the Tuskegee Syphilis Study (1932-1972), show a tendency to obey authority or follow flawed systems, even when acting against personal moral judgment. The speaker argues that the danger is not that AI will be unethical, but that its ethics will reflect our own, which are often context-dependent, biased, and inconsistent, leading to systemic harm. The ultimate test for AI ethics is whether we can embed a clear, universal ethical framework, rather than relying on culture, moment, or the flawed values embedded in training data, to guide machine decisions. Major AI companies like OpenAI, Microsoft, and Google are focusing on compliance models and fairness frameworks, but the speaker suggests the fundamental challenge is defining 'good' mathematically. The speaker quotes Mustafa Kemal Atatürk: "The ideal person is the one who uses their intellect and knowledge for the benefit of humanity."

Context: Dila Şen presents at TEDxEsade on the ethical dilemmas emerging from increasingly autonomous technologies, particularly self-driving cars and military drones. She uses thought experiments like the Trolley Problem and historical ethical failures, such as the Milgram obedience study and the Tuskegee Syphilis Study, to illustrate humanity's struggle with moral decision-making when authority or complex systems are involved. The presentation frames the challenge of AI ethics as fundamentally a human problem of defining and codifying universal values.