The surge in use of powerful generative AI technologies has raised alarms for many about the potential harm, misuse, and long-term impact on society. Many citizens are weary or even fearful of the potential damage this technology may cause, not only to users but to society as a whole. The law remains many steps behind the fast-paced development and implementation of AI systems globally, meaning that it is largely left to the discretion of companies developing AI technology to implement and adhere to ethical considerations and protect citizens from potential dangers.
As an early mover, THEMIS 5.0 has the opportunity to create an enabling pathway for trusted AI in both the public and private sector. The projects’ core values centre around improving trust in AI systems for all users and creating a strong ethical, human focus when using AI to make decisions in plethora of sectors. By creating an ecosystem that places the user at the heart of AI utility, THEMIS promises to deliver ground-breaking technology ultimately leading to consultancy, new products, jobs, and service opportunities for partners in the Consortium and beyond.
It is widely considered that bias and discrimination are two of the most significant ethical hurdles facing artificial intelligence. THEMIS 5.0 acknowledges perceptions of fairness differ among AI users (individuals, enterprises, communities, nations, cultures, societies) and may shift depending on usage of the AI system. For example, AI systems that may be non-biased are not necessarily fair as they may not accommodate for different abilities and accessibility needs. Bias can be introduced at all phases of the AI system implementation lifecycle, from the design to validation phase. THEMIS 5.0 will base its assessment of fairness bias indicators as defined by NIST, including systemic bias (in the use of AI), computational bias (in algorithms, training datasets), and human bias (in system use). The THEMIS 5.0 ‘AI-based Assessment of Fairness’ module will compute metrics for these 3 major categories of AI bias, and it will be implemented on the principles of an AI-driven anomaly detection system. Furthermore, data management in THEMIS 5.0 will be respectful of, and align with, the project’s internal ethics protocol to ensure consortium partners meet adequate ethical standards and take adequate data protection measures.
There is no question that the use of an AI tool is only as ethical as the human using it. We cannot rely on the technology to self-regulate as humans remain to be the prime operators. Megatron Transformer, an AI system developed by the Applied Deep Research team and Nvidia, when asked about its own ethics stated “AI will never be ethical. It is a tool, and like any tool, it is used for good and bad. There is no such thing as a good AI, only good and bad humans. We [the AIs] are not smart enough to make AI ethical. We are not smart enough to make AI moral…In the end, I believe that the only way to avoid an AI arms race is to have no AI at all”. This statement reiterates that despite rapid developments in AI technology, it is still the sole responsibility of humans to implement ethical frameworks into AI systems.
AI’s Ethical Concerns via Orient Software
Despite the fact that an AI system itself admits it will never be ethical, all hope is not lost. Not only are EU funded projects, such as THEMIS, pioneering a different kind of ethical technology leveraging and aiding AI, but with the implementation of The AI Act in 2024 there appears to be more emphasis on encouraging a joined-up approach to AI ethics that can be enforced across the EU. In 2020, a European ‘added value assessment’ study on a ‘European framework on ethical aspects of robotics and AI’ suggested that a common EU framework on ethics in AI (as compared with individual actions by Member States) has the potential to generate €294.9 billion in additional GDP and 4.6 million additional jobs by 2030. Cooperation among EU members could benefit the EU economy and people while also making the EU a more competitive global player. Despite the fact that many governmental and commercial entities have created standards in this field globally, there isn't a complete legal structure in place just yet. Thus, the EU can benefit from the absence of rival global regulatory governance models and fully use its position as the ‘first mover’.
Encouraging collective EU legislative action has the potential to provide European industry a competitive edge and stimulate the internal market, building on the EU's economic and regulatory authorities. Additionally, EU action would guarantee that the development, acceptance, and spread of AI is founded on the values, principles, and rights safeguarded in the EU, as well as assist the adoption of EU standards throughout the world.