Explore the vast range of titles available.

Neuroscience has focused on the detailed implementation of computation, studying neural codes, dynamics and circuits. In machine learning, however, artificial neural networks tend to eschew precisely designed codes, dynamics or circuits in favor of brute force optimization of a cost function, often using simple and relatively uniform initial architectures. Two recent developments have emerged within machine learning that create an opportunity to connect these seemingly divergent perspectives. First, structured architectures are used, including dedicated systems for attention, recursion and various forms of short- and long-term memory storage. Second, cost functions and training procedures have become more complex and are varied across layers and over time. Here we think about the brain in terms of these ideas. We hypothesize that (1) the brain optimizes cost functions, (2) the cost functions are diverse and differ across brain locations and over development, and (3) optimization operates within a pre-structured architecture matched to the computational problems posed by behavior. In support of these hypotheses, we argue that a range of implementations of credit assignment through multiple layers of neurons are compatible with our current knowledge of neural circuitry, and that the brain's specialized systems can be interpreted as enabling efficient optimization for specific problem classes. Such a heterogeneously optimized system, enabled by a series of interacting cost functions, serves to make learning data-efficient and precisely targeted to the needs of the organism. We suggest directions by which neuroscience could seek to refine and test these hypotheses.

Although a range of materials can now be fabricated using additive manufacturing techniques, these usually involve assembly of a series of stacked layers, which restricts three-dimensional (3D) geometry. Oran et al. developed a method to print a range of materials, including metals and semiconductors, inside a gel scaffold (see the Perspective by Long and Williams). When the hydrogels were dehydrated, they shrunk 10-fold, which pushed the feature sizes down to the nanoscale. Science , this issue p. 1281 ; see also p. 1244 Direct writing of nanoscale 3D structures of metals and semiconductors operates with no geometry limitations. Lithographic nanofabrication is often limited to successive fabrication of two-dimensional (2D) layers. We present a strategy for the direct assembly of 3D nanomaterials consisting of metals, semiconductors, and biomolecules arranged in virtually any 3D geometry. We used hydrogels as scaffolds for volumetric deposition of materials at defined points in space. We then optically patterned these scaffolds in three dimensions, attached one or more functional materials, and then shrank and dehydrated them in a controlled way to achieve nanoscale feature sizes in a solid substrate. We demonstrate that our process, Implosion Fabrication (ImpFab), can directly write highly conductive, 3D silver nanostructures within an acrylic scaffold via volumetric silver deposition. Using ImpFab, we achieve resolutions in the tens of nanometers and complex, non–self-supporting 3D geometries of interest for optical metamaterials.

Understanding the spatial organization of gene expression with single-nucleotide resolution requires localizing the sequences of expressed RNA transcripts within a cell in situ. Here, we describe fluorescent in situ RNA sequencing (FISSEQ), in which stably cross-linked complementary DNA (cDNA) amplicons are sequenced within a biological sample. Using 30-base reads from 8102 genes in situ, we examined RNA expression and localization in human primary fibroblasts with a simulated wound-healing assay. FISSEQ is compatible with tissue sections and whole-mount embryos and reduces the limitations of optical resolution and noisy signals on single-molecule detection. Our platform enables massively parallel detection of genetic elements, including gene transcripts and molecular barcodes, and can be used to investigate cellular phenotype, gene regulation, and environment in situ.

Neuroscience has long been an essential driver of progress in artificial intelligence (AI). We propose that to accelerate progress in AI, we must invest in fundamental research in NeuroAI. A core component of this is the embodied Turing test, which challenges AI animal models to interact with the sensorimotor world at skill levels akin to their living counterparts. The embodied Turing test shifts the focus from those capabilities like game playing and language that are especially well-developed or uniquely human to those capabilities – inherited from over 500 million years of evolution – that are shared with all animals. Building models that can pass the embodied Turing test will provide a roadmap for the next generation of AI. One of the ambitions of computational neuroscience is that we will continue to make improvements in the field of artificial intelligence that will be informed by advances in our understanding of how the brains of various species evolved to process information. To that end, here the authors propose an expanded version of the Turing test that involves embodied sensorimotor interactions with the world as a new framework for accelerating progress in artificial intelligence.

Does the brain depend on a set of elementary, reusable computations? The human cerebral cortex is central to a wide array of cognitive functions, from vision to language, reasoning, decision-making, and motor control. Yet, nearly a century after the neuroanatomical organization of the cortex was first defined, its basic logic remains unknown. One hypothesis is that cortical neurons form a single, massively repeated “canonical” circuit, characterized as a kind of a “nonlinear spatiotemporal filter with adaptive properties” ( 1 ). In this classic view, it was “assumed that these…properties are identical for all neocortical areas.” Nearly four decades later, there is still no consensus about whether such a canonical circuit exists, either in terms of its anatomical basis or its function. Likewise, there is little evidence that such uniform architectures can capture the diversity of cortical function in simple mammals, let alone characteristically human processes such as language and abstract thinking ( 2 ). Analogous software implementations in artificial intelligence (e.g., deep learning networks) have proven effective in certain pattern classification tasks, such as speech and image recognition, but likewise have made little inroads in areas such as reasoning and natural language understanding. Is the search for a single canonical cortical circuit misguided?

Current BCI technology is mainly focused on therapeutic outcomes, such as helping people with spinal-cord injuries. It might take years or even decades until BCI and other neurotechnologies are part of our daily lives. Such advances could revolutionize the treatment of many conditions, from brain injury and paralysis to epilepsy and schizophrenia, and transform human experience for the better. But the technology could also exacerbate social inequalities and offer corporations, hackers, governments or anyone else new ways to exploit and manipulate people.

Lipid membranes are key to the nanoscale compartmentalization of biological systems, but fluorescent visualization of them in intact tissues, with nanoscale precision, is challenging to do with high labeling density. Here, we report ultrastructural membrane expansion microscopy (umExM), which combines an innovative membrane label and optimized expansion microscopy protocol, to support dense labeling of membranes in tissues for nanoscale visualization. We validate the high signal-to-background ratio, and uniformity and continuity, of umExM membrane labeling in brain slices, which supports the imaging of membranes and proteins at a resolution of ~60 nm on a confocal microscope. We demonstrate the utility of umExM for the segmentation and tracing of neuronal processes, such as axons, in mouse brain tissue. Combining umExM with optical fluctuation imaging, or iterating the expansion process, yields ~35 nm resolution imaging, pointing towards the potential for electron microscopy resolution visualization of brain membranes on ordinary light microscopes. Lipid membranes are hard to visualise in tissues with nanoscale precision. The authors report ultrastructural membrane expansion microscopy (umExM), a tool that enables dense membrane labelling for nanoscale imaging of cellular membranes using a standard confocal microscope.

Simultaneously measuring the activities of all neurons in a mammalian brain at millisecond resolution is a challenge beyond the limits of existing techniques in neuroscience. Entirely new approaches may be required, motivating an analysis of the fundamental physical constraints on the problem. We outline the physical principles governing brain activity mapping using optical, electrical, magnetic resonance, and molecular modalities of neural recording. Focusing on the mouse brain, we analyze the scalability of each method, concentrating on the limitations imposed by spatiotemporal resolution, energy dissipation, and volume displacement. Based on this analysis, all existing approaches require orders of magnitude improvement in key parameters. Electrical recording is limited by the low multiplexing capacity of electrodes and their lack of intrinsic spatial resolution, optical methods are constrained by the scattering of visible light in brain tissue, magnetic resonance is hindered by the diffusion and relaxation timescales of water protons, and the implementation of molecular recording is complicated by the stochastic kinetics of enzymes. Understanding the physical limits of brain activity mapping may provide insight into opportunities for novel solutions. For example, unconventional methods for delivering electrodes may enable unprecedented numbers of recording sites, embedded optical devices could allow optical detectors to be placed within a few scattering lengths of the measured neurons, and new classes of molecularly engineered sensors might obviate cumbersome hardware architectures. We also study the physics of powering and communicating with microscale devices embedded in brain tissue and find that, while radio-frequency electromagnetic data transmission suffers from a severe power-bandwidth tradeoff, communication via infrared light or ultrasound may allow high data rates due to the possibility of spatial multiplexing. The use of embedded local recording and wireless data transmission would only be viable, however, given major improvements to the power efficiency of microelectronic devices.

We propose and theoretically study an approach to massively parallel single molecule peptide sequencing, based on single molecule measurement of the kinetics of probe binding (Havranek, et al., 2013) to the N-termini of immobilized peptides. Unlike previous proposals, this method is robust to both weak and non-specific probe-target affinities, which we demonstrate by applying the method to a range of randomized affinity matrices consisting of relatively low-quality binders. This suggests a novel principle for proteomic measurement whereby highly non-optimized sets of low-affinity binders could be applicable for protein sequencing, thus shifting the burden of amino acid identification from biomolecular design to readout. Measurement of probe occupancy times, or of time-averaged fluorescence, should allow high-accuracy determination of N-terminal amino acid identity for realistic probe sets. The time-averaged fluorescence method scales well to weakly-binding probes with dissociation constants of tens or hundreds of micromolar, and bypasses photobleaching limitations associated with other fluorescence-based approaches to protein sequencing. We argue that this method could lead to an approach with single amino acid resolution and the ability to distinguish many canonical and modified amino acids, even using highly non-optimized probe sets. This readout method should expand the design space for single molecule peptide sequencing by removing constraints on the properties of the fluorescent binding probes.

Advancements in novel neurotechnologies, such as brain computer interfaces (BCI) and neuromodulatory devices such as deep brain stimulators (DBS), will have profound implications for society and human rights. While these technologies are improving the diagnosis and treatment of mental and neurological diseases, they can also alter individual agency and estrange those using neurotechnologies from their sense of self, challenging basic notions of what it means to be human. As an international coalition of interdisciplinary scholars and practitioners, we examine these challenges and make recommendations to mitigate negative consequences that could arise from the unregulated development or application of novel neurotechnologies. We explore potential ethical challenges in four key areas: identity and agency, privacy, bias, and enhancement. To address them, we propose (1) democratic and inclusive summits to establish globally-coordinated ethical and societal guidelines for neurotechnology development and application, (2) new measures, including “Neurorights,” for data privacy, security, and consent to empower neurotechnology users’ control over their data, (3) new methods of identifying and preventing bias, and (4) the adoption of public guidelines for safe and equitable distribution of neurotechnological devices.