Stanford CS521 - AI Safety Seminar

Lecture Details

Anna Bethke is a Principal Data Scientist at Salesforce.
This talk was given on March 30, 2022.

In this session I will be discussing the components of AI Ethics, and ways that researchers, and AI practitioners can incorporate these components into every stage of their development process. These components include fairness, accountability, transparency, explainability and human rights.


Annas recommended reading materials:
https://deon.drivendata.org/
https://doteveryone.org.uk/project/consequence-scanning/
https://blog.salesforceairesearch.com/ethics-in-ai-research-papers-and-articles/

See AI programs offered at Stanford Online: https://online.stanford.edu/artificial-intelligence

0:00 Introduction
0:39 Principles of Ethical AI
3:31 Why are Af Ethics important?
4:52 Technology is Dual Use
6:29 Pose Estimation Software
7:43 What are the unintended positive and negative consequences of this technology?
18:00 What are some strategies to mitigate the negative consequences, especially the most severe ones?
22:50 Important aspects to note
25:07 Measure and Mitigate Sources of Bias
28:04 Open Source Tools
28:37 Definitions of Fairness
29:34 Algorithmic Bias Research
32:03 Dispatcher
33:30 Safety Workers
33:50 Ethical Questions We Asked
34:51 Ethical Considerations
36:26 Model Cards
37:46 Transparency Research
42:13 Stanford Classes
42:51 Salesforce Resources

#aiethics