Search

What is the prospect of developing artificial general intelligence (AGI)? I investigate this question by systematically comparing living and algorithmic systems, with a special focus on the notion of "agency." There are three fundamental differences to consider: (1) Living systems are autopoietic, that is, self-manufacturing, and therefore able to set their own intrinsic goals, while algorithms exist in a computational environment with target functions that are both provided by an external agent. (2) Living systems are embodied in the sense that there is no separation between their symbolic and physical aspects, while algorithms run on computational architectures that maximally isolate software from hardware. (3) Living systems experience a large world, in which most problems are ill-defined (and not all definable), while algorithms exist in a small world, in which all problems are well-defined. These three differences imply that living and algorithmic systems have very different capabilities and limitations. In particular, it is extremely unlikely that true AGI (beyond mere mimicry) can be developed in the current algorithmic framework of AI research. Consequently, discussions about the proper development and deployment of algorithmic tools should be shaped around the dangers and opportunities of current narrow AI, not the extremely unlikely prospect of the emergence of true agency in artificial systems.

Humanity is currently confronted with a series of profound existential crises. Beneath it all lies a crisis of meaning, of being able to make sense of the world. If we are to overcome this situation, we need robust scientific knowledge of the world and our place within it. However, science is facing a number of serious challenges to meet these needs of society. One reason for this is that science often fails to fulfil the unrealistic expectations it has set itself to meet. Such expectations are generated by an outdated view of knowledge production, which persists among many members of the public and scientists alike. This has led to an ultra-competitive system of academic research, which sacrifices long-term productivity through an excessive obsession with short-term efficiency. Efforts to diversify this system come from a movement called democratic citizen science. Here, we argue that this kind of citizen science can serve as a model for scientific inquiry in general. It requires an alternative theory of knowledge with a focus on the role that diversity plays in the process of discovery. Here, we present such an epistemology, which is based on three central philosophical pillars: perspectival realism, a naturalistic process-based epistemology, and deliberative social practices. Together, these three pillars broaden our focus from immediate research outcomes towards the cognitive and social processes which shape research strategies that facilitate sustainable long-term productivity and scientific innovation. They provide a general paradigm and template for scientific inquiry in the 21st century, which marks a shift from an industrial to an ecological vision of how scientific research should be done, and how it should be assessed. At its core are research communities that are diverse, representative, and democratic. This leads to a new, processual, paradigm of scientific project management, monitoring, and assessment, which is outlined in the last section of our paper.

Modularity is an essential feature of any adaptive complex system. Phenotypic traits are modules in the sense that they have a distinguishable structure or function, which can vary (quasi-)independently from its context. Since all phenotypic traits are the product of some underlying regulatory dynamics, the generative processes that constitute the genotype–phenotype map must also be functionally modular. Traditionally, modular processes have been identified as structural modules in regulatory networks. However, structure only constrains, but does not determine, the dynamics of a process. Here, we propose an alternative approach that decomposes the behaviour of a complex regulatory system into elementary activity-functions. Modular activities can occur in networks that show no structural modularity, making dynamical modularity more widely applicable than structural decomposition. Furthermore, the behaviour of a regulatory system closely mirrors its functional contribution to the outcome of a process, which makes dynamical modularity particularly suited for functional decomposition. We illustrate our approach with numerous examples from the study of metabolism, cellular processes, as well as development and pattern formation. We argue that dynamical modules provide a shared conceptual foundation for developmental and evolutionary biology, and serve as the foundation for a new account of process homology, which is presented in a separate contribution by DiFrisco and Jaeger to this focus issue.

We describe a collaboration between an artist, a mathematician, and a biologist, which examines the potential of drawing for understanding biological process. As a case study, it considers C. H. Waddington’s powerful visual representation of the “epigenetic landscape,” whose purpose it is to unify research in genetics, embryology, and evolutionary biology. We explore the strengths, but also the limitations of Waddington’s landscape and attempt to transcend the latter through a collaborative series of exploratory images. Through careful description of this drawing process, we touch on the epistemological consequences it had on all participants, artist and scientist alike.

Dynamic modelling is one of the cornerstones of systems biology. Many research efforts are currently being invested in the development and exploitation of large-scale kinetic models. The associated problems of parameter estimation (model calibration) and optimal experimental design are particularly challenging. The community has already developed many methods and software packages which aim to facilitate these tasks. However, there is a lack of suitable benchmark problems which allow a fair and systematic evaluation and comparison of these contributions. Here we present BioPreDyn-bench, a set of challenging parameter estimation problems which aspire to serve as reference test cases in this area. This set comprises six problems including medium and large-scale kinetic models of the bacterium E. coli, baker's yeast S. cerevisiae, the vinegar fly D. melanogaster, Chinese Hamster Ovary cells, and a generic signal transduction network. The level of description includes metabolism, transcription, signal transduction, and development. For each problem we provide (i) a basic description and formulation, (ii) implementations ready-to-run in several formats, (iii) computational results obtained with specific solvers, (iv) a basic analysis and interpretation. This suite of benchmark problems can be readily used to evaluate and compare parameter estimation methods. Further, it can also be used to build test problems for sensitivity and identifiability analysis, model reduction and optimal experimental design methods. The suite, including codes and documentation, can be freely downloaded from http://www.iim.csic.es/%7egingproc/biopredynbench/.

Insects determine their body segments in two different ways. Short-germband insects, such as the flour beetle Tribolium castaneum, use a molecular clock to establish segments sequentially. In contrast, long-germband insects, such as the vinegar fly Drosophila melanogaster, determine all segments simultaneously through a hierarchical cascade of gene regulation. Gap genes constitute the first layer of the Drosophila segmentation gene hierarchy, downstream of maternal gradients such as that of Caudal (Cad). We use data-driven mathematical modelling and phase space analysis to show that shifting gap domains in the posterior half of the Drosophila embryo are an emergent property of a robust damped oscillator mechanism, suggesting that the regulatory dynamics underlying long- and short-germband segmentation are much more similar than previously thought. In Tribolium, Cad has been proposed to modulate the frequency of the segmentation oscillator. Surprisingly, our simulations and experiments show that the shift rate of posterior gap domains is independent of maternal Cad levels in Drosophila. Our results suggest a novel evolutionary scenario for the short- to long-germband transition and help explain why this transition occurred convergently multiple times during the radiation of the holometabolan insects., Author summary Different insect species exhibit one of two distinct modes of determining their body segments (known as segmentation) during development: they either use a molecular oscillator to position segments sequentially, or they generate segments simultaneously through a hierarchical gene-regulatory cascade. The sequential mode is ancestral, while the simultaneous mode has been derived from it independently several times during evolution. In this paper, we present evidence suggesting that simultaneous segmentation also involves an oscillator in the posterior end of the embryo of the vinegar fly, Drosophila melanogaster. This surprising result indicates that both modes of segment determination are much more similar than previously thought. Such similarity provides an important step towards our understanding of the frequent evolutionary transitions observed between sequential and simultaneous segmentation.

The axial bodyplan of Drosophila melanogaster is determined during a process called morphogenesis. Shortly after fertilization, maternal bicoid mRNA is translated into Bicoid (Bcd). This protein establishes a spatially graded morphogen distribution along the anterior-posterior (AP) axis of the embryo. Bcd initiates AP axis determination by triggering expression of gap genes that subsequently regulate each other's expression to form a precisely controlled spatial distribution of gene products. Reaction-diffusion models of gap gene expression on a 1D domain have previously been used to infer complex genetic regulatory network (GRN) interactions by optimizing model parameters with respect to 1D gap gene expression data. Here we construct a finite element reaction-diffusion model with a realistic 3D geometry fit to full 3D gap gene expression data. Though gap gene products exhibit dorsal-ventral asymmetries, we discover that previously inferred gap GRNs yield qualitatively correct AP distributions on the 3D domain only when DV-symmetric initial conditions are employed. Model patterning loses qualitative agreement with experimental data when we incorporate a realistic DV-asymmetric distribution of Bcd. Further, we find that geometry alone is insufficient to account for DV-asymmetries in the final gap gene distribution. Additional GRN optimization confirms that the 3D model remains sensitive to GRN parameter perturbations. Finally, we find that incorporation of 3D data in simulation and optimization does not constrain the search space or improve optimization results.