Trustworthiness, ethics and the crucial role of open-source AI

As Artificial Intelligence (AI) becomes “smarter”, it brings forward more fundamental rather than technical problems to solve, entailing legal, ethical or even philosophical questions. In this session, the core issue of AI “alignment” is investigated under the scope of clarifying the boundaries between human and machine responsibilities, especially in the decision-making process within safety-critical applications. The current legislative status is briefly presented, primarily for the EU and USA, regarding “safe” and “trustworthy” AI, while the role of open-source models, platforms and applications is explored as probably the most important aspect. Furthermore, the moral aspects of fully autonomous AI systems are discussed through the classic “trolley problem”, demonstrating the difficulties of even formulating such tasks for designing AI. Finally, ten core principles of “aligned” AI are summarized and a few suggestions are discussed regarding how these principles may be incorporated in practice towards this direction.