As long as AI was confined to R&D laboratories, the risks of its use were limited to theoretical research topics. But the rise of this technology over the last few years and its progressive use in the professions and in the daily life of citizens have made its impact much more real and sometimes massive, especially in terms of discrimination with more or less famous proven cases.
Chatbots, voice assistants, algorithms that can handle financial transactions alone or select candidates... there are many AI-based technologies that affect society and raise questions about the ethics of these technologies.
Ethics, the keystone of trust in AI systems.
Ethics: a vast undertaking
Several initiatives are trying to establish an AI ethic. Since 2016, the number of ethics-related publications has grown exponentially.
Ethical charters are developed at several levels:
- Countries: Interactive AI geopolitical map by clicking on TAG ETHICS
- EEC: European community
- Technology services companies such as Google (principles of AI) IBM (AI ethics) Microsoft (AI principles)
- Scientific communities: Trustworthiness of AI by International Standards organisations (Joint ISO-IEC committee)
- MIT Media Lab - Harvard Berkman-Klein Center for Internet and Society
- AI foundation
- Arborus (FR)
- UNESCO (video EN / EN)
In addition to the charters, projects are launched to dive into the themes most impacted by ethical issues: justice, social or gender discrimination, AI for all, media, autonomous vehicles, health, education, etc. All of this work provides a useful framework for applying ethical principles.
Ethics: YES but how?
Put ethics at the heart of AI projects from the start.
Ask ethical questions such as:
- Does the system have sufficient volume and diversity in the input data?
- Does the system treat the criteria fairly?
- Does the system favor certain groups in the output data?
- Does the team developing the system have sufficient diversity (gender, profile, experience, skills, etc.)?
Ethics is at the heart of AI technologies. As with the safety aspects, if the team is concerned with ethics at the very end of the project, the cost of making the AI system more ethical is substantial and the result less performant. Ethical thinking in AI projects ensures alignment with the company's values. It pushes learning AI systems towards the best decision/recommendation by promoting the right balance between performance & equity.
No ethics no trust!