.png)
It was not a coincidence my path and York’s kept converging. Because of the progress in the area of AI, all of a sudden, it looked like the world was opening up to all these new possibilities. At the same time. It was opening up a whole new raft of questions, which traditional safety engineering wasn't really built to answer. I’ve always wanted to collaborate with York. John McDermid and I share a common world view, and he suggested I become a Programme Fellow, which enabled me to spend time in York collaborating with colleagues there.
Simon sees AAIP's ability to connect subject matter experts across a wide range of sectors and disciplines as key. It enables the community to get a good view of what’s going on and learn from other industries.
This wider view is increasingly important. The Programme provides the structure to look across industries to see how other sectors approach assuring safety.
There are different paradigms for safety in different sectors. Each has its own paradigm and is reaching the limits of those. Structures have become calcified in these industries – particularly in the language and setting of industry standards. The Programme can then disrupt those staid structures. It's something that the AAIP is really well placed to do because it's got this broad view and looks at the concept of autonomy itself, rather than one particular industry, and the problems associated with autonomy rather than road regulations for example.
The longer-term goal is to have safe autonomous systems with ‘functional sufficiency’ so that the system can adapt to the changing world it finds itself in.
While achieving functional sufficiency and the necessary trust of regulators and the public lies in the future, Simon sees a key achievement from the AAIP is bringing greater clarity to this landscape. The discipline of assurance of autonomous systems is evolving and beginning to take shape. AAIP is helping this by shaping the questions and defining the challenges.
I call this new field of research ‘safety assurance under uncertainty’. Structures are beginning to form whereas at the start of the AAIP it was vague, like, everything was shrouded in cloud. We are trying to understand what works, what doesn't work and how to approach these issues. The next phase, I think, is really about trying to feed this into reality.
Without the collaboration inspired and managed by the AAIP team, the likely alternative would be separate strands of research. Simon suggests that AAIP's approach works for three reasons:
[Without the AAIP] we wouldn’t have the novel, interdisciplinary ways of thinking. Instead, we would all be working independently. My work would be seen as a lone voice. The programme allows ideas to be tested and workshopped. We determine if the idea is viable, but crucially, then we have the critical mass to actually do something with it. I see a very broad, interdisciplinary consensus that can have more of an impact and achieve greater levels of acceptance.
Centre for Assuring Autonomy