Basic considerations on ethics in AI and its risk classes
In the discussion about the ethical behaviour With artificial intelligence, it is crucial to consider both the intention (= what the AI is used for) and the realisation and implementation of the technology (= how AI is developed). The type of AI used, whether it is deep learning neural networks or simple rule-based logic, is less important than the use of the AI itself.
Regulation should focus on definitional issues around acceptable uses, such as the development of better vaccines, and unacceptable uses, such as the use of AI-generated media to subtly manipulate people.
The EU's forthcoming AI Act will deal in detail with which applications are not permitted and which applications are considered particularly risky. Various risk classes are defined, each of which has different requirements. AI applications that contradict the ethics of the EU principles and pose an unacceptable risk are to be banned completely.
Data distortions, equal opportunities and transparency
In the development of AI solutions and their implementation, we should first analyse which ethical and moral aspects need to be taken into account. This includes the question of how we can deal with potential "data bias" and ensure equal opportunities, for example. This also means that potential pitfalls such as a lack of perspectives must be identified and taken into account.
While the type of AI is less relevant in terms of the "what", it does play a role in the approach ("how"). Transparency considerations are essential for people to understand what they can expect from the AI they are dealing with. This is particularly important in high-risk areas such as law enforcement or healthcare, as reflected in the AI Act.
AI ethics and general purpose AI: Intended and unintended purposes
We cannot separate the discussions about intention and development, i.e. about the why and how of AI. The latest developments in the field of "General Purpose AI (GPAI)".
GPAI refers to an AI system that is capable of performing general-purpose functions such as image and speech recognition, audio and video generation, pattern recognition, question answering, translation, etc.
It can pursue several purposes, both intended and unintended. In particular, we have recently observed many applications that use generative AI tools such as ChatGPT or Midjourney. We need all the more transparency here about how these models work and how they are used.
On the one hand, this includes transparency about the basic models themselves. Developers and regulatory authorities need this transparency in order to be able to adequately assess the use and risks of AI. On the other hand, this includes transparency about how they are used. This is relevant for users in order to understand what risks they are taking and how results are to be assessed.
Potential of Industrial AI for the economy and manufacturing industry
Artificial intelligence has a enormous potential in various economic sectors. These range from the development of new medicines to AI-controlled enzymes for decomposing plastic waste and autonomous lorry transport from hub to hub. Generative AI - AI that is able to generate text or images on demand - has extended the impact of AI into new areas such as creativity.
While we are still discovering the potential of Industrial AI, there are already concrete applications for generative AI, for example to increase productivity through the use of assistive AI and to enable hyper-personalised interactions.
When we talk about manufacturing in particular, it's easy to list the potential uses of AI, for example in the area of smart factories. It can improve operational efficiency and optimise supply chains such as warehouse and personnel processes.
However, the speed at which new technologies and paradigms are currently being introduced makes it extremely difficult to define a specific value for the economy or the manufacturing industry in general.
Protection against potential dangers from AI: Ethics & AI in dialogue
We are currently experiencing an enormous acceleration of AI transformation through the introduction of new AI technologies. This means that right now we need to prioritise the identification and mitigation of Risks pay particular attention to this. There are three levels in particular that need to be considered:
- The guarantee of fairness: Firstly, in the context of AI outcomes, we need to ensure that AI systems do not discriminate against groups of people or learn from human biases. Instead, it is important to measure and mitigate potential biases. Secondly, morality and fairness must also be ensured with regard to the availability of AI solutions: It must be ensured that the benefits are accessible to all and not just the digitally savvy part of the population.
- AI ethics and media literacy: AI will change the way we use services and how we do our work. We need to raise public awareness of how these new AI models work and what to look out for. People should be able to "read" the results of generative models, for example, and understand that the picture of the Pope with the fancy coat could be a fake, and that all the answers given by AI tools like ChatGPT do not necessarily correspond to the facts. In short, new AI literacy is needed to promote responsible use of AI, minimising potential risks while reaping the benefits of AI technology.
- Setting the course for the future of AI: Perhaps the most important topic is the question of where we want to steer our future, for example to find a responsible and moral approach to AI. What should the future look like? What are the problems we want to solve with AI in the future? What is the "AI-free zone"? And why? Now is the right time to ask ourselves these questions in order to steer the course of AI development in the right direction and ensure sufficient guard rails through regulation.
Regulation in the interests of AI and ethics: do not restrict innovation
However, the necessary regulation must not stifle innovation. Regulations should only provide guidelines for action in order to prevent risks and strengthen trust in the AI solutions developed. At the same time, they should outline a framework within which AI innovations can flourish. We therefore need appropriate and clear regulation around artificial intelligence that provides both security and allows for innovation potential.
About the person
Saara Hyvönen is one of the three co-founders of DAIN Studios, the leading AI consultancy with offices in four European countries. She holds a PhD in mathematics and has extensive experience in applying data science in both academic and business environments, for example as a postdoctoral researcher in data science at the University of Helsinki and as head of global CRM analytics at Nokia.
She specialises in data and AI strategy development, identifying optimal data use cases and defining the associated data, architecture and compliance requirements. She seeks answers to the full range of what, why and how questions. In 2021, she was listed among the 100 Brilliant Women in AI Ethics.
"I love making data work!"
Saara Hyvonen
English (UK) 
Deutsch 
Español