With a consistently growing number of new technologies that are being developed, we can argue that humankind will be capable of things even unimaginable to us now. For example, current technologies such as VR already enable us to improve the efficiency of workflows and AI is giving us tools to create art by simply typing a query. However, it is difficult to manage how people will make use of new technologies. There will always be people with malicious intent, unfortunately.
That is where Safety by Design (SbD) can come in as part of the solution. Safety by Design is a relatively new design principle where new inventions are designed from the start with user safety in mind. With SbD, safety is not only a small component but a prerequisite for the completion of the development process. This prevents unwanted behavior and outcomes (INHOPE, 2021). It forces more thought on possible uses and minimizes threats to users (World Economic Forum, n.d.). It fits in the larger concept of responsible innovation, which is centered on ethical acceptability, sustainability, and societal desirability, according to Von Schomberg (2013).
One example where Safety by design has been implemented is DALL-E 2. This creative AI system, based on GPT-3, has recently been released to the public in batches. However, it has disabled any functions with regard to the creation of unwanted output, such as violence, explicit content, or other of such topics, even before the tool was released (OpenAI, n.d.-a; OpenAI, n.d.-b). The developers behind DALL-E 2 have done extensive research on possible implications and possible uses their system has (Ahmad & Mischkin, 2022).
Safety by Design is already being applied in the development of physical products, and although rising in use, it is not as conventional in technology development (Perrino, 2022). Should it be applied more often by developers? As Safety by design does keep society safe from those with bad intentions, it does raise the question of where the border lies between freedom and belittlement. Who determines what is right or wrong, what constitutes as malicious behavior, and most importantly: will it stand in the way of innovation?
- Ahmad, L. & P. Mischkin (2022). DALL-E 2 Preview – Risks and Limitations. Accessed via https://github.com/openai/dalle-2-preview/blob/main/system-card.md#probes-and-evaluations.
- INHOPE (2021). What is Safety by Design? Accessed via https://inhope.org/EN/articles/what-is-safety-by-design.
- OpenAI (n.d.-a). Content policy. Accessed via https://labs.openai.com/policies/content-policy.
- OpenAI (n.d.-b). DALL-E 2. Accessed via https://openai.com/dall-e-2/.
- Von Schomberg, R. 2013. A vision of responsible research and innovation. In Responsible innovation, ed. R. Owen, M. Heintz, and J. Bessant. London: Wiley.
- World Economic Forum (n.d.). Safety by Design (SbD). Accessed via https://www.weforum.org/projects/safety-by-design-sbd.
