Course Summary

1. Course Summary

You've completed this course, Responsible AI for Developers: Privacy and Safety. Let's recap what you have learned. In this course, we introduced privacy in machine learning and some best practices on privacy. We described two approaches to help developers achieve privacy in training data, de-identify your data, and randomize your data. To de-identify your data, you can apply redaction, replacement, masking, tokenization, bucketing, or shifting. To randomize your data, you can apply data perturbation or differential privacy. It is recommended to apply the most suitable technique based on your business requirements for data privacy. We introduced two approaches to help developers achieve privacy in machine learning model training, Differentially Private Stochastic Gradient Descent, also known as DP-SGD, and federated learning. You also learned about system security best practices in Google Cloud. Focus on sensitive data protection with the Cloud Data Loss Prevention (DLP) API, Encryption with Key Management Service (KMS), Access control with Identity and Access Management (IAM), and Monitoring. We also introduced AI safety. You learned about several major failure modes and adversarial testing techniques for safety evaluation. You identified what are safety classifiers and what are input, output safeguards that help you prevent harm. You learned about two approaches, instruction-tuning, and reinforcement learning from human feedback, to teach safety to AI models. You also learned the safety settings in Gemini and Natural Language API. As artificial intelligence continues its rapid ascent, the conversation around responsible AI becomes ever more vital. New technological developments constantly present fresh challenges and opportunities in this domain. It's even more important now to ensure that when you develop for AI, you are equipped with the latest insights and best practices for responsible AI implementation.

2. Let's practice!

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.