Ethical AI, Bias and Machine Learning

Presented at DevFest Trichy 2022 hosted by GDG Trichy, India, the presentation is about understanding ethical issues in AI, the bias, and the ways of doing machine learning in an ethical manner

Transcript

1. Trichy Wesley Kambale ML Engineer Ethical AI, Bias and Machine Learning
2. What’s Machine Learning?
3. the use of data and algorithms to imitate intelligent human behavior
6. Classification
7. Regression
8. Clustering
9. Sequence Prediction
10. Style Transfer
11. Collect Data Train & Test Model Define Objective Predict & Evaluate Focus on User Machine Learning Process
12. Bias in AI…
13. Occurs when two data sets are not considered equal
14. - AI-based recruiting tool showing bias against women - AI-based Twitter accounts tweeting racist comments - Facial recognition tool exhibiting bias toward certain ethnicities
16. Basic AI Process
17. Internal Factors - No focus on bias identification - Non-diverse Teams - Non-identification of sensitive data attributes and related correlations - Unclear policies
18. External Factors - Biased real-world data - No frameworks for bias identification - Biased third-party AI Systems
19. Ethical AI…
20. Mitigation - Entity-level controls - Establish AI governance and policies - Promote a culture of ethics and diversity - Prepare balanced data sets
21. Further Mitigation - Account for bias in AI modeling - Make periodic assessments - Build explainable AI
22. Conclusion AI systems are not equal in terms of bias risk. For example, an AI system that suggests products for a shopping cart has less risk than an AI system that determines whether to approve an individual’s job application.
23. Trichy Wesley Kambale ML Engineer Thank you!