Innovation Management Module 7 Trust in AI
Arms race in AI
Ai security vs skill of adversarial attack
Main Limitations of AI
Bias Adversarial attack
Why does AI bias matter
‒ Hiring tool that discriminated against women ‒ Facial recognition working better for light-skinned than dark-skinned individuals ‒ Bank loan approvals ‒ Dangerous effect of reinforcing unhealthy stereotype
How to generate trust in AI
‒ Importance of data ‒ Correct selection ‒ High quality data ‒ Maintenance to counter model drift ‒ Building the right system ‒ Trade-off between simple systems and complex deep learning-based systems (?) ‒ Black boxes ‒ System design choices ‒ Cognitive and emotional trust ‒ Technical capabilities and form of AI representation
How to combat biases
‒ Technical solutions ‒ E.g., "zero out" the bias in words ‒ Use less biased and / or more inclusive data ‒ Transparency and / or auditing processes ‒ Diverse workforce ‒ Creates less biased applications
Goldilocks Rule
‒ Too optimistic: Sentient / super-intelligent AI Killer robots coming soon ‒ Too pessimistic: AI cannot do everything Another AI winter is coming ‒ Just right: AI can't do everything But it will transform industries