Suchergebnisse

Generative artificial intelligence models present the EU with regulatory, global governance, and security dilemmas. Brussels should work with partners to mitigate risks and set norms for trustworthy AI.

Enabled by recent improvements in generation methodologies, DeepFakes have become mainstream due to their increasingly better visual quality, the increase in easy-to-use generation tools and the rapid dissemination through social media. This fact poses a severe threat to our societies with the potential to erode social cohesion and influence our democracies. To mitigate the threat, numerous DeepFake detection schemes have been introduced in the literature but very few provide a web service that can be used in the wild. In this paper, we introduce the MeVer DeepFake detection service, a web service detecting deep learning manipulations in images and video. We present the design and implementation of the proposed processing pipeline that involves a model ensemble scheme, and we endow the service with a model card for transparency. Experimental results show that our service performs robustly on the three benchmark datasets while being vulnerable to Adversarial Attacks. Finally, we outline our experience and lessons learned when deploying a research system into production in the hopes that it will be useful to other academic and industry teams. ; This work has been supported by the AI4Media H2020 project, partially funded by the European Commission under contract number 951911, and the US Paris Tech Challenge award, which was funded by the US Department of State Global Engagement Center under contract number SGECPD18CA0024.

AbstractEthics of technology systems have become an area of interest in academic research as well as international policy in recent years. Several organisation have consequently published principles of ethical artificial intelligence (AI) in line with this trend. The documents identify principles, values, and other abstract requirements for AI development and deployment. Critics raise concerns about whether these documents are in fact constructive, or if they are produced as a higher form of virtue signalling. A theme that is beginning to become apparent in the academic literature regarding these documents is the inherent lack of effective and practical methods and processes for producing ethical AI. This article attempts a critical analysis which draws upon ethical AI documents from a range of contexts including company, organisational, governmental, and academic perspectives. Both the theoretical and practical components of AI guidelines are explored and analysed, consequently bringing to light the necessity of introducing a measurable component to such documents for the purpose of ensuring a positive outcome of deploying AI systems based on ethical principles. We propose a minimal framework for stakeholders to develop AI in an ethical and human-centred manner.

Rapidly developing Artificial Intelligence (AI) systems hold tremendous potential to change various domains and exert considerable influence on societies and organizations alike. More than merely a technical discipline, AI requires interaction between various professions. Based on the results of fundamental literature and empirical research, this book addresses the management's awareness of the ethical and moral aspects of AI. It seeks to fill a literature gap and offer the management guidance on tackling Trustworthy AI Implementation (TAII) while also considering ethical dependencies within the company. The TAII Framework introduced here pursues a holistic approach to identifying systemic ethical relationships within the company ecosystem and considers corporate values, business models, and common goods aspects like the Sustainable Development Goals and the Universal Declaration of Human Rights. Further, it provides guidance on the implementation of AI ethics in organisations without requiring a deeper background in philosophy and considers the social impacts outside of the software and data engineering setting. Depending on the respective legal context or area of application, the TAII Framework can be adapted and used with a range of regulations and ethical principles. This book can serve as a case study or self-review for c-level managers and students who are interested in this field. It also offers valuable guidelines and perspectives for policymakers looking to pursue an ethical approach to AI

The European approach to artificial intelligence (AI) points to ethical considerations, human control and trustworthiness as its core tenets. But how clearly is this approach reflected in the Member States' strategies?This anthology analyses to what extent the notions of ethical and trustworthy AI, presented by the High-Level Expert Group on Artificial Intelligence and the European Commission, have influenced AI strategies in Portugal, The Netherlands, Italy, the Czech Republic, Poland, Norway as well as the Nordics overall.It is clear that the EU-level policies have had an impact on the national level strategies, although sometimes only to the extent that they were published before the national documents. For instance, while some countries, such as Norway and Portugal, have explicitly incorporated aspects from the Ethics Guidelines, others, such as the Nordics, already tended to include questions of trust and transparency, or on ethics as in the case of Poland.The EU has emphasised AI trustworthiness as both an ethical imperative and competitive advantage. However, implementation is still at the starting line: much depends on alignment between this diverse group of nations, with different priorities, within the single market.

Während es inzwischen eine ganze Reihe praktischer Leitfäden für die Implementierung des Konzepts der vertrauenswürdigen künstlichen Intelligenz (KI) gibt, fehlt es an konkreten Beispielen und Projekten für Umsetzungen, anhand derer sich Probleme und Erfolgsstrategien der Akteur*innen vor Ort aufzeigen ließen. Dieser Beitrag stellt ausgewählte Umsetzungsprojekte vor. Durchweg zeigt sich dabei ein noch geringer Grad an Konkretisierung. Deshalb wird anschließend nach den Gründen für das Umsetzungsdefizit gefragt. Drei Erklärungen kommen infrage: Time-to-Market-Überlegungen aufseiten der Unternehmen, Unklarheit darüber, welche Aspekte des Konzepts der vertrauenswürdigen KI bei welchen Anwendungen überhaupt relevant sind sowie die Tatsache, dass die Umsetzung von KI‑Projekten komplexer ist als die Umsetzung 'normaler' Software-Projekte und deshalb spezifische Vorkehrungen notwendig sind.

AbstractDue to the extensive progress of research in artificial intelligence (AI) as well as its deployment and application, the public debate on AI systems has also gained momentum in recent years. With the publication of the Ethics Guidelines for Trustworthy AI (2019), notions of trust and trustworthiness gained particular attention within AI ethics-debates; despite an apparent consensus that AI should be trustworthy, it is less clear what trust and trustworthiness entail in the field of AI. In this paper, I give a detailed overview on the notion of trust employed in AI Ethics Guidelines thus far. Based on that, I assess their overlaps and their omissions from the perspective of practical philosophy. I argue that, currently, AI ethics tends to overload the notion of trustworthiness. It thus runs the risk of becoming a buzzword that cannot be operationalized into a working concept for AI research. What is needed, however, is an approach that is also informed with findings of the research on trust in other fields, for instance, in social sciences and humanities, especially in the field of practical philosophy. This paper is intended as a step in this direction.

The full benefit from using AI to generate value for businesses, societal wellbeing and the environment is still to be fully realised. To lower adoption barriers of Industrial AI, challenges on multiple levels (technical complexity, trustworthiness, industrialisation, data frameworks and infrastructures, etc.) need to be addressed to benefit from its full socio economic potential for economy, society and welfare. It is now time to foster the development of "industrial and trustworthy AI" and nurture European innovation and sovereignty ambitions and benefit society and leading European industries. In this position paper, a comprehensive industrial and trustworthy AI framework that clusters the priority area for AI research, innovation and deployment is introduced. It covers tools and methodologies that support the design, test, validation, verification, and maintainability of AI based functions and systems and addresses the development of AI based process and systems to demonstrate its integration into new products and services. Conformity assessment schemes, balancing innovation, business and European perspectives, are considered to connect risk management, functional andtrustworthiness requirements to industrial processes. In addition adequate standards supporting industrial AI and trustworthiness will play a central role. Implementing the industrial and trustworthy AI framework will require resources beyond the means of any European private stakeholders. Therefore, strong support from ecosystems, governments, and Europe might not be an option but necessary.