“There’s no continuum. Current claims and hopes for progress in models for making computers intelligent are like the belief that someone climbing a tree is making progress toward reaching the moon.” – Stuart Dreyfus, Mind over machine
After several decades of evolutionary computation, and 20 years of the journal Genetic Programming and Evolvable Machines (GPEM), what’s the state of the field? Are we making meaningful progress towards being able to automatically evolve software, structures, and machines whose performance and behavior matters? Are these developments leading us in important directions? How does our work relate to the high profile successes in other areas of machine learning and artificial intelligence? What is our roadmap going forward as a field?
Are we making meaningful discoveries, or just climbing trees to reach the moon?
To celebrate 20 years of GPEM we aim to assemble an anniversary issue that is worthy of the fine work in the first two decades of the journal, and address some of these important questions. Our goal is both to take stock and to look ahead; to combine rigorous and diagnostic overviews with sharp new ideas that have the potential to challenge and reshape the field.
Like any publishing venture, we want people to read and refer to this issue for some time to come. How can that best happen? We believe through a combination of:
- High quality review articles. A good review article can be an enormous service to the field, and receive numerous citations down the road. We want more than just an organized list of citations with limited commentary; this provides little more value than a good Internet search, and will age rapidly. We’re looking for surveys that go beyond the “hits”, make valuable connections, and provide insightful analysis, helping us better understand who we are and what we’re doing.
- Challenge pieces. A piece (even a quite short one) that challenges the field in a meaningful way can be an important spur for action. This could be a “grand challenge” style article (which, to be honest, is probably a lot harder for a still young field like ours than it was for Hilbert), a more focussed challenge (e.g., “modularity really matters – here’s why and how we’d know we were making progress”), or a more methodological challenge (like recent work on benchmarking in GP, or challenges on how we best compare our work to other “hot” areas like deep neural networks). Challenges should explain why progress on this challenge matters, and provide meaningful ways to gauge progress or success; careful comparison of promising tools and techniques would also have value.
- Groundbreaking new research. This is arguably the hardest category to judge; we all hope that our work is remarkable and will revolutionize the field, but most work is incremental, and the volume of publishing ensures that much of it will have limited impact. If you have work that you believe has the potential to profoundly affect the field, however, this would be a particularly apt venue, as anniversary issues have the potential for more mentions, search hits, etc., giving you a leg up in your quest for world domination. We’re looking here for research that could really change how we think about the field and our work; if you feel you have those kinds of ideas, please submit.
Our submission deadline is 17 Oct 2018, and we hope to have initial reviewing done by 14 Nov 2018. Resubmissions will be due 1 Dec 2018, with final notifications made by 9 Jan 2019. Questions, etc., should be sent to Nic McPhee at gpem20th@gmail.com
Looking forward to receiving some amazing work!
The official Call for Papers on the Springer site is here.
The official Call for Papers on the Springer site is here.
