Building a POC or black box solution is relatively easy
but building AI solutions we can trust is extremely
complicated. The risk of not knowing how your solution
made a decision, if it is reliable, secure, and fair is
extremely high. The consequences of bad AI can range from
reputational risk, loss of IP, lawsuits, financial loss,
and much more. For AI to be Trustworthy it needs to have
at least 5 additional properties to the overall accuracy:
- Explainability
- Fairness
- Robustness
- Transparency
- Privacy