Explore the potential pitfalls and dangers of AI systems in autonomous decision-making through this insightful conference talk. Delve into critical issues such as bias, fairness, confounding variables, adversarial attacks, ethics, and explainability. Gain a high-level understanding of the security and privacy concerns surrounding AI applications for individuals and society. Learn from computer scientist Joachim Ganseman as he shares his expertise on topics including natural language processing, conversational interfaces, and their applications in government and social security administration. Discover the implications of AI in various contexts, from phishing attempts and fake news to recommendation systems and conspiracy theories. Examine the importance of explainable AI, awareness, and policy-making in addressing these challenges, with a focus on European Union initiatives.
Overview
Syllabus
Intro
What do you do
Whats the interest
Whats the fear
What can go wrong
Build your own AI
Bias
No production
Confounding factors
Semantic gap
Bias fairness
Machine learning systems
Unexpected behavior
Data poisoning
adversarial examples
robust attacks
stickers on objects
curse of dimensionality
other examples
good AI systems
phishing attempts
social media
Phishing
LinkedIn
Fake news
Fake images
Fake text
Recommendation systems
YouTube
Conspiration theories
YouTube recommendation system
Fake hoaxes
Explainable AI
Scams
Awareness
GDPR
Policymakers
The EU
Smallboots
Taught by
Devoxx
