Skip to content

Conclusions of the workshops

Here we present the conclusions drawn on each of the Evobody workshops,


Workshop 1

The first Evobody workshop aimed to obtain an overview of the most basic aspects of embodied evolution: definition, feasibility and milestones, as well as desirability and usefulness issues.

Even though no single definition of embodied evolution could be agreed upon, the most important aspect of embodiment seems to be that there is no separation between the information and the material that processes that information. In itself, this represents a revolutionary step, for instance, towards personal computers that become integrated into everyday environments. Besides, the goal of embodied evolution is not to understand evolution, but to successfully use its principles and methods to obtain the computational systems of the future.

However, in order to obtain non-biological organisms (a mix of hardware, software and wetware components) that can perform complex computations and, in fact, evolve into even more advanced systems, a lot still needs to happen. For instance, if one consider reproduction as a essential part of evolutionary systems, great developments have to be made in the area of smart material engineering in order to obtain (self-)replicating non-biological systems. Moreover, if control over designed systems is to be guaranteed, appropriate methodologies that allow embodied systems’ design and implementation have to be established.

Existing dis-embodied methods, such as simulation and evolutionary computing, can be used as a design toolbox and remain relevant to the understanding of general properties of evolutionary systems, model and predict the behaviour of systems.

In this context, the interaction and cross-fertilisation between ICT, bio-techonology and material science is one of the major challenges. It is from these that the effective realisation of embodied evolution will come: the role of each discipline needs to be identified, the objectives and complexity of systems evaluated, and issues such as trust, controllability, reliability and safety discussed.


Workshop 2

The second EVOBODY workshop had the objective of engaging deeper in the discussion of (some of) the main challenges related to embodied evolution, namely body types, reproduction, kill switch & self-repair, useful applications, and online design methodology. Here we provide a short overview on each topic.

Embodied artificial evolution (EAE) requires a body. With respect to the design of the body, two main streams have been identified: (i) biochemical, and (ii) mechatronic/robotic. Biochemical systems have a clear advantage over mechatronic devices, as biological properties, such as reproduction, can be taken for granted. In the mechatronic level we miss the properties of living organisms, but we gain crucial advantages such as programmability, modularity and portability.

One major challenge of EAE systems is the creation of new units/individuals. The important concept is the production of a new object of the same type, where new individuals should potentially have increasing functional complexity and heritable genetic material. The transfer of inheritable genetic material between individuals could speed up evolution in many settings.

(Self-)Reproduction allows the information about a species to be preserved even though individuals die, while self-repair allows the structure of the individual to be maintained despite flaws arising in component parts. At different levels of analysis, both can be seen as part of an information-preserving process. It can be noted that self-repair is currently a hot topic in material science generally. This has not yet reached robotics, even though desirable.

In terms of design methodologies, the most prominent transitional change caused by EAE technology is that design will not end with manufacturing. In EAE systems, design and manufacturing become an online, continuous, intertwined activity, fuelled by the evolutionary operators. This empowers systems to be autonomous and self-improving. The unprecedented challenge is an on-the-fly monitoring and steering of systems, in line with the given/desired user preferences.

A serious concern regarding EAE is the possibility of ‘runaway evolution’. Runaway evolution as we use it here stands for the process of uncontrolled population growth. Such a growth might also be accompanied by the emergence of new, unwanted features in the population. As technology becomes more alive, the risk of the ‘runaway evolution’ might increase. It is, therefore, crucial to be truthful about what we are doing and maybe – after a few generations – a kill switch will be needed. However, the overall agreement is that no real danger is imminent.

In terms of applications of EAE involving biotechnology, i.e. at microscopic level, one possible application is to evolve drugs that adapt to the characteristics/needs of a particular patient. At macroscopic level, a field of application of EAE concerns evolutionary robotics. Search and rescue robots with evolving morphology to adapt their geometry for (cooperative) transportation is an example. In general, there has been an agreement on the necessity of physical morphology and control to co-evolve in order to have meaningful applications.

As a summary, we note that the integration of bio-chemistry, micro-biology, synthetic biology, and robotics will be a vital challenge in coming years and promises several radical breakthroughs regarding the adaptive, developmental and evolutionary properties of artificial systems.

Advertisements
%d bloggers like this: