L6: OpenAI

• data protection

• privacy

• transparency

• accountability

• biases

Privacy: appropriate ethical level of data privacy collected; privacy input (data for training) + privacy output (predictions about people). The extent to which people have knowledge and control over their data without undue pressure.

Explainability: the space between the inputs and the outputs, the extent you can precisely assess what has been going on

Bias: coming from real-world discriminations, under- sampling, benchmark bias, wrong metrics, etc.

The goal is to leverage powerful AI tools to achieve a desirable social impact, focusing on responsible business practices and sustainable AI development.

Ethical risk mitigation involves a comprehensive and all-encompassing approach, focusing on understanding the risks and operationalizing ethical risk mitigation strategies.

The principles are:

Beneficence

Non-maleficence

Autonomy, Justice

Explicability

They focus on promoting well-being, preventing risks, balancing decision power, ensuring fairness, and maintaining transparency and accountability.

AI contributes by enabling self-realization without devaluing human abilities, enhancing human agency without removing responsibility, increasing societal capabilities without reducing human control, and cultivating societal cohesion without eroding self-determination.

Considerations include biases and fairness, privacy and cybersecurity, impacts on jobs and work nature, compliance with legal frameworks, misuse of AI, and environmental impacts.

The alignment problem involves embedding human values in machines and ensuring that machines are developed and used according to organizational values.

Key components include governance structure, ethical content, clear values, cross-organizational awareness, product development standards, internal accountability, tracking standards adherence, and control mechanisms.

The model involves understanding AI tech workings, assessing reliability levels, transparency about risks, complementing AI to increase reliability, and continuous control and improvement processes.