Adaptive collective systems are we ready to let go of control?

The field of computer science is rapidly changing, and we often barely seem to notice it. For years we have been working on the same topics, and many of us expect to continue do so for still a long time to come. Moore’s law is no longer determining the speedups of our programs: new algorithms and insights are pushing us further to an extent that some believe we need to start worrying about Artificial Intelligence. Computer science is eating the world, and we are the ones responsible for that. Yet, we want to stay in control. I argue that we need get into a different mindset: let go of control. A while back, I joined a team of experts on computational intelligence in an attempt to write a booklet on adaptive collective systems. It taught me a lot about how colleagues in the same field were looking at what is important in computer science. Coming from computer systems research, it now seems to me that my view is conservative. Many colleagues concentrating on developing software constructs are often being conservative as well. We need to let a few things go. I will talk about adaptive collective systems as being a special type of distributed computer system in which control, or rather, the lack of control, plays a crucial role. Control is no longer fixed; it evolves through learning mechanisms. The take-away message is that our future distributed systems need to be architected with learning facilities. However, considering that there are so many different types of computer systems, it is not obvious how learning should be incorporated.