Eric Mitchell

Bio: Eric Mitchell is an academic researcher from Samsung. The author has contributed to research in topics: Computer science & Biology. The author has an hindex of 6, co-authored 22 publications receiving 153 citations. Previous affiliations of Eric Mitchell include Stanford University & Princeton University.

FlyWire: Online community for whole-brain connectomics

Abstract: Due to advances in automated image acquisition and analysis, new whole-brain connectomes beyond C. elegans are finally on the horizon. Proofreading of whole-brain automated reconstructions will require many person-years of effort, due to the huge volumes of data involved. Here we present FlyWire, an online community for proofreading neural circuits in a fly brain, and explain how its computational and social structures are organized to scale up to whole-brain connectomics. Browser-based 3D interactive segmentation by collaborative editing of a spatially chunked supervoxel graph makes it possible to distribute proofreading to individuals located virtually anywhere in the world. Information in the edit history is programmatically accessible for a variety of uses such as estimating proofreading accuracy or building incentive systems. An open community accelerates proofreading by recruiting more participants, and accelerates scientific discovery by requiring information sharing. We demonstrate how FlyWire enables circuit analysis by reconstructing and analysing the connectome of mechanosensory neurons.

Functional connectomics spanning multiple areas of mouse visual cortex

Abstract: The value of an integrated approach for understanding the neocortex by combining functional characterization of single neuron activity with the underlying circuit architecture has been understood since the dawn of modern neuroscience. However, in practice, anatomical connectivity and physiology have been studied mostly separately. Following in the footsteps of previous studies that have combined physiology and anatomy in the same tissue, here we present a unique functional connectomics dataset that contains calcium imaging of an estimated 75,000 neurons from primary visual cortex (VISp) and three higher visual areas (VISrl, VISal and VISlm), that were recorded while a mouse viewed natural movies and parametric stimuli. The functional data were co-registered with electron microscopy (EM) data of the same volume which were automatically segmented, reconstructing more than 200,000 cells (neuronal and non-neuronal) and 524 million synapses. Subsequent proofreading of some neurons in this volume yielded reconstructions that include complete dendritic trees as well the local and inter-areal axonal projections. The largest proofread excitatory axon reached a length of 19 mm and formed 1893 synapses, while the largest inhibitory axon formed 10081 synapses. Here we release this dataset as an open access resource to the scientific community including a set of analysis tools that allows easy data access, both programmatically and through a web user interface.

On the Opportunities and Risks of Foundation Models.

Abstract: AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.

DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature

Abstract: The fluency and factual knowledge of large language models (LLMs) heightens the need for corresponding systems to detect whether a piece of text is machine-written. For example, students may use LLMs to complete written assignments, leaving instructors unable to accurately assess student learning. In this paper, we first demonstrate that text sampled from an LLM tends to occupy negative curvature regions of the model's log probability function. Leveraging this observation, we then define a new curvature-based criterion for judging if a passage is generated from a given LLM. This approach, which we call DetectGPT, does not require training a separate classifier, collecting a dataset of real or generated passages, or explicitly watermarking generated text. It uses only log probabilities computed by the model of interest and random perturbations of the passage from another generic pre-trained language model (e.g, T5). We find DetectGPT is more discriminative than existing zero-shot methods for model sample detection, notably improving detection of fake news articles generated by 20B parameter GPT-NeoX from 0.81 AUROC for the strongest zero-shot baseline to 0.95 AUROC for DetectGPT. See https://ericmitchell.ai/detectgpt for code, data, and other project information.

Offline Meta-Reinforcement Learning with Advantage Weighting

Abstract: This paper introduces the offline meta-reinforcement learning (offline meta-RL) problem setting and proposes an algorithm that performs well in this setting. Offline meta-RL is analogous to the widely successful supervised learning strategy of pre-training a model on a large batch of fixed, pre-collected data (possibly from various tasks) and fine-tuning the model to a new task with relatively little data. That is, in offline meta-RL, we meta-train on fixed, pre-collected data from several tasks in order to adapt to a new task with a very small amount (less than 5 trajectories) of data from the new task. By nature of being offline, algorithms for offline meta-RL can utilize the largest possible pool of training data available and eliminate potentially unsafe or costly data collection during meta-training. This setting inherits the challenges of offline RL, but it differs significantly because offline RL does not generally consider a) transfer to new tasks or b) limited data from the test task, both of which we face in offline meta-RL. Targeting the offline meta-RL setting, we propose Meta-Actor Critic with Advantage Weighting (MACAW), an optimization-based meta-learning algorithm that uses simple, supervised regression objectives for both the inner and outer loop of meta-training. On offline variants of common meta-RL benchmarks, we empirically find that this approach enables fully offline meta-reinforcement learning and achieves notable gains over prior methods.

A connectomic study of a petascale fragment of human cerebral cortex

Abstract: We acquired a rapidly preserved human surgical sample from the temporal lobe of the cerebral cortex. We stained a 1 mm3 volume with heavy metals, embedded it in resin, cut more than 5000 slices at ~30 nm and imaged these sections using a high-speed multibeam scanning electron microscope. We used computational methods to render the three-dimensional structure containing 57,216 cells, hundreds of millions of neurites and 133.7 million synaptic connections. The 1.4 petabyte electron microscopy volume, the segmented cells, cell parts, blood vessels, myelin, inhibitory and excitatory synapses, and 104 manually proofread cells are available to peruse online. Many interesting and unusual features were evident in this dataset. Glia outnumbered neurons 2:1 and oligodendrocytes were the most common cell type in the volume. Excitatory spiny neurons comprised 69% of the neuronal population, and excitatory synapses also were in the majority (76%). The synaptic drive onto spiny neurons was biased more strongly toward excitation (70%) than was the case for inhibitory interneurons (48%). Despite incompleteness of the automated segmentation caused by split and merge errors, we could automatically generate (and then validate) connections between most of the excitatory and inhibitory neuron types both within and between layers. In studying these neurons we found that deep layer excitatory cell types can be classified into new subsets, based on structural and connectivity differences, and that chandelier interneurons not only innervate excitatory neuron initial segments as previously described, but also each others initial segments. Furthermore, among the thousands of weak connections established on each neuron, there exist rarer highly powerful axonal inputs that establish multi-synaptic contacts (up to ~20 synapses) with target neurons. Our analysis indicates that these strong inputs are specific, and allow small numbers of axons to have an outsized role in the activity of some of their postsynaptic partners.

Quantifying behavior to understand the brain.

Abstract: Over the past years, numerous methods have emerged to automate the quantification of animal behavior at a resolution not previously imaginable. This has opened up a new field of computational ethology and will, in the near future, make it possible to quantify in near completeness what an animal is doing as it navigates its environment. The importance of improving the techniques with which we characterize behavior is reflected in the emerging recognition that understanding behavior is an essential (or even prerequisite) step to pursuing neuroscience questions. The use of these methods, however, is not limited to studying behavior in the wild or in strictly ethological settings. Modern tools for behavioral quantification can be applied to the full gamut of approaches that have historically been used to link brain to behavior, from psychophysics to cognitive tasks, augmenting those measurements with rich descriptions of how animals navigate those tasks. Here we review recent technical advances in quantifying behavior, particularly in methods for tracking animal motion and characterizing the structure of those dynamics. We discuss open challenges that remain for behavioral quantification and highlight promising future directions, with a strong emphasis on emerging approaches in deep learning, the core technology that has enabled the markedly rapid pace of progress of this field. We then discuss how quantitative descriptions of behavior can be leveraged to connect brain activity with animal movements, with the ultimate goal of resolving the relationship between neural circuits, cognitive processes and behavior.

Galactica: A Large Language Model for Science

Abstract: Information overload is a major obstacle to scientific progress. The explosive growth in scientific literature and data has made it ever harder to discover useful insights in a large mass of information. Today scientific knowledge is accessed through search engines, but they are unable to organize scientific knowledge alone. In this paper we introduce Galactica: a large language model that can store, combine and reason about scientific knowledge. We train on a large scientific corpus of papers, reference material, knowledge bases and many other sources. We outperform existing models on a range of scientific tasks. On technical knowledge probes such as LaTeX equations, Galactica outperforms the latest GPT-3 by 68.2% versus 49.0%. Galactica also performs well on reasoning, outperforming Chinchilla on mathematical MMLU by 41.3% to 35.7%, and PaLM 540B on MATH with a score of 20.4% versus 8.8%. It also sets a new state-of-the-art on downstream tasks such as PubMedQA and MedMCQA dev of 77.6% and 52.9%. And despite not being trained on a general corpus, Galactica outperforms BLOOM and OPT-175B on BIG-bench. We believe these results demonstrate the potential for language models as a new interface for science. We open source the model for the benefit of the scientific community1.

GLM-130B: An Open Bilingual Pre-trained Model

Abstract: We introduce GLM-130B, a bilingual (English and Chinese) pre-trained language model with 130 billion parameters. It is an attempt to open-source a 100B-scale model at least as good as GPT-3 and unveil how models of such a scale can be successfully pre-trained. Over the course of this effort, we face numerous unexpected technical and engineering challenges, particularly on loss spikes and disconvergence. In this paper, we introduce the training process of GLM-130B including its design choices, training strategies for both efficiency and stability, and engineering efforts. The resultant GLM-130B model offers significant outperformance over GPT-3 175B on a wide range of popular English benchmarks while the performance advantage is not observed in OPT-175B and BLOOM-176B. It also consistently and significantly outperforms ERNIE TITAN 3.0 260B -- the largest Chinese language model -- across related benchmarks. Finally, we leverage a unique scaling property of GLM-130B to reach INT4 quantization, without quantization aware training and with almost no performance loss, making it the first among 100B-scale models. More importantly, the property allows its effective inference on 4$\times$RTX 3090 (24G) or 8$\times$RTX 2080 Ti (11G) GPUs, the most ever affordable GPUs required for using 100B-scale models. The GLM-130B model weights are publicly accessible and its code, training logs, related toolkit, and lessons learned are open-sourced at https://github.com/THUDM/GLM-130B .