Evolution sculpts both the body plans and nervous systems of agents together over time. In contrast, in AI and robotics, a robot’s body plan is usually designed by hand, and control policies are then optimized for that fixed design. The task of simultaneously co-optimizing the morphology and controller of an embodied robot has remained a challenge — as evidenced by the little improvement upon early techniques over the decades since their introduction. Embodied cognition posits that behavior arises from a close coupling between body plan and sensorimotor control, which suggests why co-optimizing these two subsystems is so difficult: most evolutionary changes to morphology tend to adversely impact sensorimotor control, leading to an overall decrease in behavioral performance. Here, we further examine this hypothesis and demonstrate a technique for “morphological innovation protection”, which reduces selection pressure on recently morphologically-changed individuals, thus enabling evolution some time to “readapt” to the new morphology with subsequent control policy mutations. This treatment tends to yield individuals that are significantly more fit than those that existed before the morphological change and increases evolvability. We also show the potential for this method to avoid local optima and show fitness increases further into optimization, as well as the potential for convergence to similar highly fit morphologies across widely varying initial conditions. While this technique is admittedly only the first of many steps that must be taken to achieve scalable optimization of embodied machines, we hope that theoretical insight into the cause of evolutionary stagnation in current methods will help to enable the automation of robot design and behavioral training.
Scalable Co-Optimization of Morphology and Control in Embodied Machines
Nick Cheney, Josh Bongard, Vytas SunSpiral, Hod Lipson
For Isaac Newton, laying the foundation of modern physics and astronomy was a bit of a sideshow. He believed that his truly important work was deciphering ancient scriptures and uncovering the nature of the Christian religion. True, his skill in calculation was helpful for describing celestial mechanics, but far more critical was applying it to Hebrew prophecies.
How do we think about his career when we consider that Newton wrote vastly more on religious subjects than he did on what we would consider scientific ones? Rob Iliffe’s new book Priest of Nature pulls back the curtain on what Newton thought of as his life’s work, rather than that for which we remember him.
The enlightened empiricist
Priest of Nature: The Religious Worlds of Isaac Newton Rob Iliffe Oxford University Press, 2017. 536 pp.
Science 30 Jun 2017:
Vol. 356, Issue 6345, pp. 1341
Despite the growth of Open Access, potentially illegally circumventing paywalls to access scholarly publications is becoming a more mainstream phenomenon. The web service Sci-Hub is amongst the biggest facilitators of this, offering free access to around 62 million publications. So far it is not well studied how and why its users are accessing publications through Sci-Hub. By utilizing the recently released corpus of Sci-Hub and comparing it to the data of ~28 million downloads done through the service, this study tries to address some of these questions. The comparative analysis shows that both the usage and complete corpus is largely made up of recently published articles, with users disproportionately favoring newer articles and 35% of downloaded articles being published after 2013. These results hint that embargo periods before publications become Open Access are frequently circumnavigated using Guerilla Open Access approaches like Sci-Hub. On a journal level, the downloads show a bias towards some scholarly disciplines, especially Chemistry, suggesting increased barriers to access for these. Comparing the use and corpus on a publisher level, it becomes clear that only 11% of publishers are highly requested in comparison to the baseline frequency, while 45% of all publishers are significantly less accessed than expected. Despite this, the oligopoly of publishers is even more remarkable on the level of content consumption, with 80% of all downloads being published through only 9 publishers. All of this suggests that Sci-Hub is used by different populations and for a number of different reasons, and that there is still a lack of access to the published scientific record. A further analysis of these openly available data resources will undoubtedly be valuable for the investigation of academic publishing.
Looking into Pandora’s Box: The Content of Sci-Hub and its Usage
Stochastic growth processes give rise to diverse and intricate structures everywhere in nature, often referred to as fractals. In general, these complex structures reflect the non-trivial competition among the interactions that generate them. In particular, the paradigmatic Laplacian-growth model exhibits a characteristic fractal to non-fractal morphological transition as the non-linear effects of its growth dynamics increase. So far, a complete scaling theory for this type of transitions, as well as a general analytical description for their fractal dimensions have been lacking. In this work, we show that despite the enormous variety of shapes, these morphological transitions have clear universal scaling characteristics. Using a statistical approach to fundamental particle-cluster aggregation, we introduce two non-trivial fractal to non-fractal transitions that capture all the main features of fractal growth. By analyzing the respective clusters, in addition to constructing a dynamical model for their fractal dimension, we show that they are well described by a general dimensionality function regardless of their space symmetry-breaking mechanism, including the Laplacian case itself. Moreover, under the appropriate variable transformation this description is universal, i.e., independent of the transition dynamics, the initial cluster configuration, and the embedding Euclidean space.
Universal fractality of morphological transitions in stochastic growth processes
J. R. Nicolás-Carlock, J. L. Carrillo-Estrada & V. Dossetti
Scientific Reports 7, Article number: 3523 (2017)