Lecture Notes: Decision-Making Frameworks and Ethics in AI
Introduction
Speaker shares personal childhood experiences with bedtime stories and nursery rhymes.
Nursery rhymes in Bengali often convey historical or wisdom insights.
Example: "A good crop of mangoes is a signal for a good yield of rice" and "Abundance of tamarind is a warning for an impending flood."
Historical Context of Decision-Making
Ancestors' Practices:
Utilized observation to develop farming heuristics and predictions.
Developed strategies and tactics in hunting, requiring memory and cognition.
Decisions that led to survival were seen as successful.
Evolution of Decision-Making
Today's AI and Machine Learning:
Form generalizations based on data (similar to ancient heuristics).
Machine learning applications span various fields: medical, financial, agricultural, etc.
However, data biases lead to flawed recommendations.
Case Studies: Risks of Current AI Systems
Anne Hathaway Effect:
AI sentiment analysis linked to stock market trends, causing unintended consequences.
Gender and Racial Bias:
Boston University study shows gender biases in AI.
Bias in AI leads to problematic decisions, e.g., US courts’ use of COMPAS software:
Higher false positive rates for black defendants.
Highlights systemic biases in data-driven predictions.
The Role of Ethics in AI
Need for Ethical Consideration:
AI systems must integrate ethical standards to avoid perpetuating past biases.
Decisions should not solely rely on statistical correctness.
Religion as a Parallel:
Religion offers an ethical framework but has failed when it harms individuals.
Long-term effective decision-making frameworks require fairness and equitability.
Conclusion
Frameworks for Decision-Making:
Effective frameworks must withstand the test of time and adhere to ethical standards.
AI must prioritize harm mitigation and ethics.
The importance of balancing efficiency with serving individuals in society.
Final Thoughts:
Emphasize on "having all the facts but also the wisdom to know which ones to apply."
Key Takeaway:
Ethical considerations are crucial in developing and applying AI and machine learning. Without them, modern tools may not survive as past traditional wisdom has.