Why AI Isn’t Fully Reliable Yet
AI can make dangerous errors, especially in high-stakes areas like medical diagnosis
AI model failed at predicting 5-year breast/lung cancer outcomes → major blind spots.
What Is Superintelligence & Why Are Experts Worried?
Superintelligence = AI reaching/surpassing human-level intelligence.
Example: Thought to be decades away, but experts now warn it may arrive much sooner and could become uncontrollable.
What Does the Superintelligence Statement Call For?
A ban on developing superintelligence until:
- It is scientifically proven safe
- The public strongly agrees
Harari warns it could “break the operating system of human civilization.”
Autonomy vs. Freedom — Why Autonomy Matters More
Autonomy = self-determination in action; freedom is too messy/conflicted.
COVID mask debates show how “freedom” is unclear, but autonomy focuses on meaningful choice.
What Is Authentic Autonomy?
Authentic autonomy = choices that truly come from you, not from algorithmic nudging.
If Instagram only shows certain lifestyles/careers, it subtly shapes what you think your “choices” are.
Autonomy ≠ “Do Whatever You Want”
Autonomy is socially grounded, not individualistic; your sense of self forms through communities, culture, language, school.
Your decisions reflect family, culture, hobbies — not pure independence.
Minimal Integrity of the Self
Your “self” must remain bounded and not dissolved into algorithmic systems.
If algorithms shrink your “domain of possibility” (due to money, time, targeted nudges), autonomy declines.
Two Types of Harm in Data Colonialism
Answer:
- Actual harm (data used against you)
Ex: Biased algorithms denying loans, insurance, jobs. - Possibility of harm (panopticon effect)
Ex: Self-censoring online because “it stays forever.”
Datafication threatens four core human rights values:
Autonomy → constant tracking removes choice
Dignity → Amazon workers timed so tightly they skip bathroom breaks
Equality → biased parole/credit algorithms
Community → social data is extracted and sold
Wong’s 4 properties of data:
- Shared (others create data about you – tags, messages)
- Mundane (collected automatically from everyday behavior)
- Linked (copied, backed up, shared, aggregated)
- Forever (nearly impossible to fully delete)
