1
Q
Interpretability
A
how accurate a ML model can associate a cause to an effect
2
Q
Explainability
A
ability of the parameters in justifying the results
3
Q
Specific examples of risks in AI ?
A
- bias (mostly human)
- LLM hallucination (solution: COSTAR)
- Backdoors/malware on ML (difficult to find)
- Deep fakes (fake videos and photos)
- Phishing
4
Q
AI and security
A
- can improve security
- Attackers can use AI to improve attacks
- AI is target of the attack
5
Q
New risks in AI
A
Transparency
bias
accuracy/hallucination
security
data protection
legal concerns
6
Q
Global impact
A
- whole humnaity
- create global competitive disadvantage
- co intelligence
7
Q
Audit of AI
A
identify relevance of AI within company
- AI strategy
- inventory of usage
- capability assessment
8
Q
Introduce audit within AI lifecycle
A
design
developement
deployment
monitoring
GRC
flashcards
Decks in class (5)
# Cards
