Artificial intelligence and systems engineering
Self-driving cars need artificial intelligence (AI), and after several vehicle manufacturers have already followed Google and developed self-driving cars, now Toyota are included in this group.
One problem I see with regard to SE is that the system should be deterministic, especially in embedded systems: Given the same input, the same output should come. This should ensure comprehensible behavior, which is often considered essential for functional safety, among other things. In systems engineering, unpredictable behavior is also referred to as "emergent behavior". This is a classic sign of complex (as opposed to complicated) systems.
This is easier said than done with AI systems: the input (e.g. the map material) changes slowly, and the inputs generated in traffic are so plentiful that deterministic behavior would not necessarily ensure greater safety. Apart from that, one of the characteristics of an AI system is its ability to learn. For determinism, this would have to be switched off, which would drastically reduce its value.
Photo: Dr. Gill Pratt, Toyota Executive Technical Advisor and Chief Executive Officer of Toyota Research Institute, Inc.
We simply have to stop trying to control everything. The world is not predictable: It's better to simply adapt to it.
Where can I find details on the difference between complex and complicated systems?
Many SE books go into this in detail, such as the INCOSE manual [1] (also available in German). But briefly explained using an example: A clock is complicated, you can understand exactly how it works by looking at the individual parts. Road traffic is complex: The traffic develops characteristics (e.g. traffic jams, etc.) that are not apparent from the individual part (car) (this is then „emergent behavior").
[1] http://www.incose.org/ProductsPublications/sehandbook