NeuroLinx Research Institute

Since the advent of widely accessible AI tools, AI technology has been in high demand by businesses, academic researchers and individuals. Technology companies are building AI infrastructure at a rapid pace, and these facilities consume vast and growing resources, particularly electricity and water, with significant real and projected climate impacts. There is a need for new research initiatives to support long time horizon efforts to develop energy efficient computing capabilities to support the continued growth of AI infrastructure in a sustainable fashion. Such efficiency is required at both the hardware and software levels. Where can industry turn for examples of ultra-low power, energy efficient computing? We argue here that neurobiological principles offer rich and under-exploited sources of inspiration for energy efficient NeuroAI, and that new partnerships between industry and academia should be developed in this direction.

We used computer simulations of growth, mating and death of cephalopods and fishes to explore the effect of different life-history strategies on the relative prevalence of alternative male mating strategies. Specifically, we investigated the consequences of single or multiple matings per lifetime, mating strategy switching, cannibalism, resource stochasticity, and altruism towards relatives. We found that a combination of single (semelparous) matings, cannibalism and an absence of mating strategy changes in one lifetime led to a more strictly partitioned parameter space, with a reduced region where the two mating strategies co-exist in similar numbers. Explicitly including Hamilton's rule in simulations of the social system of a Cichlid led to an increase of dominant males, at the expense of both sneakers and dwarf males ("super-sneakers"). Our predictions provide general bounds on the viable ratios of alternative male mating strategies with different life-histories, and under possibly rapidly changing ecological situations.

<p>We argue that the high energy use by present-day semiconductor computing technology will prevent the emergence of an artificial intelligence system that could reasonably be described as a “superintelligence”. This hard limit on artificial superintelligence (ASI) emerges from the energy requirements of a system that would be more intelligent but orders of magnitude less efficient in energy use than human brains. An ASI would have to supersede not only a single brain, but a large population of humans, further multiplying the energy requirement. A hypothetical ASI would likely consume orders of magnitude more energy than what is available in industrialized society. We estimate the energy use by ASI with an equation we term the ”Erasi equation”, for the <em>E</em>nergy <em>R</em>equirement for <em>A</em>rtificial <em>S</em>uper<em>I</em>ntelligence. Additional efficiency consequences will emerge from the current unfocussed and scattered developmental trajectory of AI research. Taken together, these arguments suggest that the emergence of an ASI is highly unlikely in the foreseeable future based on current computer architectures, primarily due to energy constraints, with biomimicry being a possible solution.</p>

Abstract We used computer simulations of growth, mating and death of cephalopods and fishes to explore the effect of different life-history strategies on the relative prevalence of alternative male mating strategies. Specifically, we investigated the consequences of single or multiple matings per lifetime, mating strategy switching, cannibalism, resource stochasticity, and altruism towards relatives. We found that a combination of single (semelparous) matings, cannibalism and an absence of mating strategy changes in one lifetime led to a more strictly partitioned parameter space, with a reduced region where the two mating strategies co-exist in similar numbers.Explicitly including Hamilton’s rule in simulations of the social system of a Cichlid led to an increase of dominant males, at the expense of both sneakers and dwarf males (“super-sneakers”). Our predictions provide general bounds on the viable ratios of alternative male mating strategies with different life-histories, and under possibly rapidly changing ecological situations.

Fish escape from approaching threats via a stereotyped escape behavior. This behavior, and the underlying neural circuit organized around the Mauthner cell command neurons, have both been extensively investigated experimentally, mainly in two laboratory model organisms, the goldfish and the zebrafish. However, fish biodiversity is enormous, a number of variants of the basal escape behavior exist. In marine gobies (a family of small benthic fishes) which share burrows with alpheid shrimp, the escape behavior has likely been partially modified into a tactile communication system which allow the fish to communicate the approach of a predatory fish to the shrimp. In this communication system, the goby responds to intermediate-strength threats with a brief tail-flick which the shrimp senses with its antennae.We investigated the shrimp goby escape and communication system with computational models. We asked how the circuitry of the basal escape behavior could be modified to produce behavior akin to the shrimp-goby communication system. In a simple model, we found that mutual inhibitions between Mauthner cells can be tuned to produce an oscillatory response to intermediate strength inputs, albeit only in a narrow parameter range.Using a more detailed model, we found that two modifications of the fish locomotor system transform it into a model reproducing the shrimp goby behavior. These modifications are: 1. modifying the central pattern generator which drives swimming such that it is quiescent when receiving no inputs; 2. introducing a direct sensory input to this central pattern generator, bypassing the Mauthner cells.

Abstract We conducted a visual and photographic survey of the gobiidae in the Bolinao area of the Philipines, located on the western tip of the Lingayen gulf, on the west coast of Luzon island. We identified a total of 40 species, of which 18 are shrimp-associated. One species found (Myersina lachneri) constitutes a range expansion into the Philippines. This number of species is in the expected range compared to other studies of marine goby faunae in the coral triangle, despite the significant anthropogenic pressures onto the marine ecosystem in the surveyed area.

Neuromorphic engineering is the approach to intelligent machine design inspired by nature. Here, we outline possible robotic design principles derived from the neural and motor systems of sea urchins (Echinoida). Firstly, we review the neurobiology and locomotor systems of sea urchins, with a comparative emphasis on differences to animals with a more centralized nervous system. We discuss the functioning and enervation of the tube feet, pedicellariae, and spines, including the limited autonomy of these structures. We outline the design principles behind the sea urchin nervous system. We discuss the current approaches of adapting these principles to robotics, such as sucker-like structures inspired by tube feet and a robotic adaptation of the sea urchin jaw, as well as future directions and possible limitations to using these principles in robots.

Abstract We analyze the number of marine fish species as a function of fish body size and occurrence depth. For this purpose, we analyze the FishBase database. We compare these data to predictions of fish species numbers derived from the neutral theory of biodiversity combined with well-established ecological scaling laws, and measured oceanic biomass data. We consider several variants of these scaling laws, and we find that more large fish species exist compared to the prediction, which is especially true for elasmobranchs, possibly due to their overwhelmingly predatory niches. We find species numbers decreasing with occurrence depth somewhat quicker than our predictions based on the decrease of the number of individuals with depth indicates. This is especially true for the elasmobranchs. This is unsurprising, since the individuals versus depth data did not specifically determine elasmobranch biomass, and since sharks are known to be limited to depths < 3,000 m.Finally, we discuss how a reduced rate of speciation in larger animals could explain why large species are rare, in spite of the advantages of large body sizes outlined in Cope’s rule.