Ethical Pathway for Safe Artificial Intelligence (AI) with Five Pillars

1st Pillar: University Courses on Ethics

Today, the computer community is not well acquainted with ethical and moral principles. Therefore, the first step of this proposal is the enhancement of this knowledge at the University level through specific, whole-semester mandatory courses. The aim of the course could be to understand ethical principles and their application to computer science and to prepare the students to make ethical and moral decisions in their professional careers.

The course may include definitions of ethics, overviews of ethics theories and their evolutions over the years, professional ethics, ethical considerations in software development, distributions, and transparencies, the significant impacts of AI on human and society's behavior, cybersecurity, and effects of automation, among others.

2nd Pillar: An Oath Like the Hippocratic for Computer Scientists

An oath, similar to the Hippocratic one for medical students, was suggested for AI scientists.[12] This oath using simple and understandable ethical principles introduces moral practices into the AI community.

The aim and benefits of such an oath could summarized:

• Improve the ethical standards in AI science by preventing corrupt practices: hacking, data poisoning, cyber-war, and others.
• Increase the awareness of the potentially lethal dangers of AI.
• Make clear the personal responsibilities for unethical activities, their consequences, and the potential personal penalties.
• Prevent AI involved in medical malpractice and the production of lethal autonomous, nuclear, and biological weapons.
• Increase global collaborations to develop solid systems for safe AI.[12,13,14]

3rd Pillar: Globally Accepted Ethical Rules for AI

Although the founders of computer science, Allen Turing in the 40s and Wiener in the 60s had noticed the ethical implications and threats of AI, only recently International Organizations and Governments paid attention to those dangers.[15]

Today, there are several publications on ethical issues, principles, and rules for the use of AI but none have reached global acceptance. Moreover, these efforts had more theoretical merit without specifying ways for their implementation. It is obvious, that before any Global acceptance is achieved, significant financial, social, cultural, ethical, and legal issues have to be solved.

The more recent and comprehensive document of AI ethics is the one produced by the European Union in2019 and its revision published in 2024.[16] This publication covers most of the current issues of AI involvement and focuses on business fairness, transparency, and legality.[16]

However, a global treaty urgently needed to be developed at the highest authoritarian level, such as the one of the United Nations, and signed by all its country members.

It is considered that if the majority (if not all) of the countries of the world agree, the chances of the global efforts to develop a safe and solidly friendly AI towards humanity could be more easily accomplished.

The treaty should include rules that AI should be used by all humans with fairness, equality, privacy, transparency, and accountability.

Furthermore, the principal aim of the treaty should be to ensure the safe and beneficial use of AI for humanity, in the future. In that case, the treaty should promote the global collaboration of AI scientists in developing the proper algorithms to ensure a safe and human-friendly AI.

4th Pillar: Ethical Committees for AI

The implementation of the Global Treaty needs local Ethical Committees to be established.

These committees at the universities, Research centers, Scientific Societies, Publication Institutes, Computing Companies, and Organizations will have the task of verifying the ethical standards and merits of each algorithm produced in-house before its publication or becoming commercially available. This procedure minimizes errors that may happen during the race of the major AI companies ( Google, Microsoft, Intel, Chat-AI, and others) to develop, the most advanced, accurate, fast, and cheap AI system of the market. It is obvious, that the above companies have to establish their internal ethical Committees.

Thus, the global treaty should also cover the way that the local ethical committees would function, especially those acting for commercial corporations. Overall, the goals of these committees should be the protection and privacy of the data, its quality and integrity, its transparency and availability, as well as, data rights and accountability.

5th Pillar: Embed ethical principles into AI

Theoretically, this pillar of the pathway could be developed in two stages: one to verify the most proper principles of digital ethics and a second these principles to be technically embedded into machines, by AI engineers.

It is apparent, that the first stage is an extremely difficult one because the task of producing a globally accepted set of ethical rules is enormous. It will have to overcome very strong obstacles concerning philosophical, legal, cultural, religious, and ethical issues.

Although extremely difficult, all these problems have to be solved well on time, with compromises and wisdom.

The second phase of this pillar will be to embed the globally accepted modes of ethics into the functions of AI and its proper algorithms.

This may achieved by expert computer engineers using already existing technologies like machine learning and automatic self-improving techniques or those invented in the future. The task could be a ‘machine’ more ethical than the humans.