LTAT.06.028 Advanced Trustworthy AI
Are your AI systems non-discriminatory, transparent, and resilient? This technical course goes beyond a general understanding of what is at stake and introduces the more practical methods for implementing trustworthy AI. Every developer working with AI needs these skills.
If you are new in topic, please consider taking first the foundational course, Trustworthy AI
- Lectures: MOOC (Online-based)
- Course coordinator: Huber Flores
Length: Four-week program Target audience: Software developers, data scientists, engineering leads
Part I: Trustworthy AI in organisations and industrial environments
- Trustworthy AI requirements in practice
- Developing trustworthy AI in an organisation
Part II: Bias and fairness in AI
- Sources of bias in the evaluation methods for fairness
- Methods for assessing and documenting model bias and fairness
- Tools for evaluating fairness in your AI model
Part III: Dissecting the internal logic of machine learning
- Glass-box and black-box - What’s the difference?
- Instrumenting AI models with XAI
- Applied explainable AI
- Interpreting outputs
Part IV: Resilience AI: Defence from security and privacy attacks
- Attacks against machine learning systems
- Technical solutions to mitigate attacks
- Industry examples of privacy and security
Part V: Conclusions
- Interactive case: trustworthiness of service recommendation AI
- Epilogue
Announcements
- Certificate for UT students and staff is available for free, please present your certificate of passing the course.
