Explainability concern requires judicious use of AI

Significance

This presents a safety concern in some areas, and creates difficulties in guaranteeing that AI decisions are unbiased, for example in their treatment of different demographic groups.

Impacts

• Adoption of deep learning AI in areas such as law, finance and employment will raise concerns over bias and discrimination.
• Lack of explainability represents a potential safety concern in areas such as medicine and autonomous vehicles.
• Inexplicable deep learning AI could be manipulated by malicious actors, creating a security vulnerability in defence applications.