Search for: All records

The emergence of seemingly similar representations across tasks and neural architectures suggests that convergent properties may underlie sophisticated behavior. One form of representation that seems particularly fundamental to reasoning in many artificial (and perhaps natural) networks is the formation of world models, which decompose observed task structures into re-usable perceptual primitives and task-relevant relations. In this work, we show that auto-regressive transformers tasked with solving mazes learn to linearly represent the structure of mazes, and that the formation of these representations coincides with a sharp increase in generalization performance. Furthermore, we find preliminary evidence for Adjacency Heads which may play a role in computing valid paths through mazes. 

Book Chapter in © Springer Nature Switzerland AG 2019 T. J. McGenity (ed.), Microbial Communities Utilizing Hydrocarbons and Lipids: Members, Metagenomics and Ecophysiology, Handbook of Hydrocarbon and Lipid Microbiology, https://doi.org/10.1007/978-3-030-14785-3_12 

The Nebraska Innovation Maker Co-Laboratory (NiMC) project is developing a model to establish and support makerspaces in rural communities. As makerspaces gain in popularity a chasm has developing between urban access and lack thereof for rural populations. The MiMC model supports collaboration between university faculty and staff and rural makerspaces by utilizing virtual reality and telepresence robots. The exploratory research project deployed telepresence robotics to teach, co-teach and provide project support to a rural community (pop. 7,000) makerspace. Virtual reality was used to teach creativity concepts, VR digital creation and digital to physical manifestation of projects. The NiMC project will continue to explore the model of connection and support of rural makerspaces. 

Abstract During aerobic oxidation of methane (CH₄) in seawater, a process which mitigates atmospheric emissions, the¹²C‐isotopologue reacts with a slightly greater rate constant than the¹³C‐isotopologue, leaving the residual CH₄isotopically fractionated. Prior studies have attempted to exploit this systematic isotopic fractionation from methane oxidation to quantify the extent that a CH₄pool has been oxidized in seawater. However, cultivation‐based studies have suggested that isotopic fractionation fundamentally changes as a microbial population blooms in response to an influx of reactive substrates. Using a systematic mesocosm incubation study with recently collected seawater, here we investigate the fundamental isotopic kinetics of aerobic CH₄oxidation during a microbial bloom. As detailed in a companion paper, seawater samples were collected from seep fields in Hudson Canyon, U.S. Atlantic Margin, and atop Woolsey Mound (also known as Sleeping Dragon) which is part of lease block MC118 in the northern Gulf of Mexico, and used in these investigations. The results from both Hudson Canyon and MC118 show that in these natural environments isotopic fraction for CH₄oxidation follows a first‐order kinetic process. The results also show that the isotopic fractionation factor remains constant during this methanotrophic bloom once rapid CH₄oxidation begins and that the magnitude of the fractionation factor appears correlated with the first‐order reaction rate constant. These findings greatly simplify the use of natural stable isotope changes in CH₄to assess the extent that CH₄is oxidized in seawater following seafloor release. 

Abstract Microbial aerobic oxidation is known to be a significant sink of marine methane (CH₄), contributing to the relatively minor atmospheric release of this greenhouse gas over vast stretches of the ocean. However, the chemical kinetics of aerobic CH₄oxidation are not well established, making it difficult to predict and assess the extent that CH₄is oxidized in seawater following seafloor release. Here we investigate the kinetics of aerobic CH₄oxidation using mesocosm incubations of fresh seawater samples collected from seep fields in Hudson Canyon, U.S. Atlantic Margin and MC118, Gulf of Mexico to gain a fundamental chemical understanding of this CH₄sink. The goals of this investigation were to determine the response or lag time following CH₄release until more rapid oxidation begins, the reaction order, and the stoichiometry of reactants utilized (i.e., CH₄, oxygen, nitrate, phosphate, trace metals) during CH₄oxidation. The results for both Hudson Canyon and MC118 environments show that CH₄oxidation rates sharply increased within less than one month following the CH₄inoculation of seawater. However, the exact temporal characteristics of this more rapid CH₄oxidation varied based on location, possibly dependent on the local circulation and biogeochemical conditions at the point of seawater collection. The data further suggest that methane oxidation behaves as a first‐order kinetic process and that the reaction rate constant remains constant once rapid CH₄oxidation begins.