Good AI or Bad AI? How to Decode Responsible Artificial Intelligence

Artificial intelligence or AI has become increasingly powerful and prevalent in recent years. While AI technologies like ChatGPT, DALL-E, and other AI platforms provide exciting new capabilities, they also raise important ethical concerns about bias, privacy, security, and safety.

To help ensure AI benefits society and minimizes harm, researchers have developed responsible AI principles. These include human and environmental well-being, human-centered values, fairness, privacy protection, reliability, transparency, contestability, and accountability.

Examples of irresponsible AI include biased facial recognition, privacy violations, generating misinformation, and unreliable systems. To achieve the best possible AI future, users should monitor AI systems and report issues.

Responsible AI means developing and using AI in a way that provides benefits while minimizing risks. According to researchers at CSIRO, responsible AI principles include:

Human and environmental well-being: AI should benefit individuals, society and the environment. For example, AI for medical diagnosis or environmental protection.

Human-centered values: AI should respect human rights, diversity and autonomy. Transparent and explainable AI should be embedded with human values.

Fairness: AI should be inclusive, accessible and avoid unfair discrimination. For example, some facial recognition has been criticized for bias against people of color and women.

Privacy protection and security: AI should respect and uphold privacy rights. Personal data should only be collected when necessary and properly safeguarded. Some AI like Clearview AI has violated privacy laws.

Reliability and safety: AI should reliably operate as intended. Pilot studies can help avoid issues like the chatbot Tay generating hate speech.

Transparency and explainability: The use of AI should be transparent and clearly disclosed. People should understand AI impacts and limitations, e.g. chatbots can hallucinate.

Contestability: There should be ways to challenge the use or outcomes of AI that significantly impacts a person, group or environment.

Accountability: People responsible for AI should be identifiable and accountable, with human oversight of AI systems. Look for AI developed by those who promote ethical, responsible practices.

You can spot irresponsible AI by looking for violations of these principles like lack of transparency or unfair outcomes. Report issues to providers and authorities to build a better future with AI.