Lecture Notes: The Impact of AI on the Future

Introduction to AI's Impact

Difference in Task Assignment:
- Humans understand context and implied limits.
- AI systems are given fixed objectives that need to be explicitly defined.
Problem with AI Objectives:
- AI systems don't factor in unspoken common sense constraints.
- Example: Fixing ocean acidification could inadvertently harm the atmosphere.
Solution:
- AI needs to recognize its own limitations in understanding objectives.
- Machines should be designed to exhibit uncertainty about objectives.
- Control over AI derives from its uncertainty, avoiding rigid "psychopathic" behaviors.
- Similar parallels exist with human behavior.

Technological Unemployment:
- Concept dates back to Aristotle, further explored by Keynes as "technological unemployment."
- AI could replace human labor as it advances.
Current Automation Examples:
- Partially automated warehouses in e-commerce.
- Full automation could displace millions of jobs.

Dependence on Machines:
- Risks of losing human skills and understanding.
- E.M. Forster's story analogy: Civilization managed by machines leads to a loss of human capability.
- "WALL-E" serves as a modern example.
Continuation of Knowledge:
- Civilization continuity depends on unbroken chains of teaching and learning.
- Risk of the chain breaking as AI advances.

Arrival and Impact:
- No specific date for general purpose AI's arrival.
- Impact will gradually increase with AI advancements.
Predictions:
- Experts estimate general purpose AI by end of the century.
- Median prediction around 2045; some predict a longer timeline.
- John McAfee humorously indicates a broad timeline from 5 to 500 years.

Development of general purpose AI is complex, requiring significant breakthroughs.
Continuous exploration and understanding are necessary to harness AI responsibly.