Search

This paper investigates the application of the transformer architecture in protein folding, as exemplified by DeepMind's AlphaFold project, and its implications for the understanding of large language models as models of language. The prevailing discourse often assumes a ready-made analogy between proteins -- encoded as sequences of amino acids -- and natural language -- encoded as sequences of discrete symbols. Instead of assuming as given the linguistic structure of proteins, we critically evaluate this analogy to assess the kind of knowledge-making afforded by the transformer architecture. We first trace the analogy's emergence and historical development, carving out the influence of structural linguistics on structural biology beginning in the mid-20th century. We then examine three often overlooked pre-processing steps essential to the transformer architecture, including subword tokenization, word embedding, and positional encoding, to demonstrate its regime of representation based on continuous, high-dimensional vector spaces, which departs from the discrete, semantically demarcated symbols of language. The successful deployment of transformers in protein folding, we argue, discloses what we consider a non-linguistic approach to token processing intrinsic to the architecture. We contend that through this non-linguistic processing, the transformer architecture carves out unique epistemological territory and produces a new class of knowledge, distinct from established domains. We contend that our search for intelligent machines has to begin with the shape, rather than the place, of intelligence. Consequently, the emerging field of critical AI studies should take methodological inspiration from the history of science in its quest to conceptualize the contributions of artificial intelligence to knowledge-making, within and beyond the domain-specific sciences.

Over the past decade, cryogenic electron microscopy (cryo-EM) has emerged as a primary method for determining near-native, near-atomic resolution 3D structures of biological macromolecules. In order to meet increasing demand for cryo-EM, automated methods to improve throughput and efficiency while lowering costs are needed. Currently, all high-magnification cryo-EM data collection softwares require human input and manual tuning of parameters. Expert operators must navigate low- and medium-magnification images to find good high-magnification collection locations. Automating this is non-trivial: the images suffer from low signal-to-noise ratio and are affected by a range of experimental parameters that can differ for each collection session. Here, we use various computer vision algorithms, including mixture models, convolutional neural networks, and U-Nets to develop the first pipeline to automate low- and medium-magnification targeting. Learned models in this pipeline are trained on a large internal dataset of images from real world cryo-EM data collection sessions, labeled with locations that were selected by operators. Using these models, we show that we can effectively detect and classify regions of interest in low- and medium-magnification images, and can generalize to unseen sessions, as well as to images captured using different microscopes from external facilities. We expect our open-source pipeline, Ptolemy, will be both immediately useful as a tool for automation of cryo-EM data collection, and serve as a foundation for future advanced methods for efficient and automated cryo-EM microscopy., Comment: Main: 12 pages, 11 figures. Appendix: 2 pages, 1 figure

Micro and nanomechanical resonators with ultra-low dissipation have great potential as useful quantum resources. The superfluid micromechanical resonators presented here possess several advantageous characteristics: straightforward thermalization, dissipationless flow, and in situ tunability. We identify and quantitatively model the various dissipation mechanisms in two resonators, one fabricated from borosilicate glass and one from single crystal quartz. As the resonators are cryogenically cooled into the superfluid state, the damping from thermal effects and from the normal fluid component are strongly suppressed. At our lowest temperatures, damping is limited solely by internal dissipation in the substrate materials, and reach quality factors up to 913,000 at 13 mK. By lifting this limitation through substrate material choice and resonator design, modelling suggests that the resonators should reach quality factors as high as 10$^8$ at 100 mK, putting this architecture in an ideal position to harness mechanical quantum effects., Comment: 25 pages, 12 figures, 3 tables

On-chip cavity optomechanics, in which strong co-localization of light and mechanical motion is engineered, relies on efficient coupling of light both into and out of the on-chip optical resonator. Here we detail our particular style of tapered and dimpled optical fibers, pioneered by the Painter group at Caltech, which are a versatile and reliable solution to efficient on-chip coupling. First, a brief overview of tapered, single mode fibers is presented, in which the single mode cutoff diameter is highlighted. The apparatus used to create a dimpled tapered fiber is then described, followed by a comprehensive account of the procedure by which a dimpled tapered fiber is produced and mounted in our system. The custom-built optical access vacuum chambers in which our on-chip optomechanical measurements are performed are then discussed. Finally, the process by which our optomechanical devices are fabricated and the method by which we explore their optical and mechanical properties is explained. It is our expectation that this manuscript will enable the novice to develop advanced optomechanical experiments., Comment: 31 pages, 9 figures

Existing literature in educational technology and media has tended to overlook the larger role educational institutions play as socializing forces, which instill children with national identities and values. Recognizing the important role schools play in forming children's characters and perspectives on social problems, this article advocates a research agenda that focuses on how educational technologies serve as tools to combat larger social issues such as peace, health, and poverty through innovative approaches to traditional curricula and schooling. Focusing specifically on the ongoing Israeli-Palestinian conflict, this article specifies a model for global, mobile stories that strive to develop cross-border awareness and promote a sense of global identity. This study presents findings from the collection of 185 stories from Palestinian youth, and suggests avenues for future research at the intersection of education, technology, and innovative approaches to solving social problems. [For the complete volume, "Educational Media and Technology Yearbook, Volume 36, 2011," see ED597737.]

Literacy is a human right unequally distributed among the world's population. Despite global efforts to fight illiteracy, high illiteracy rates continue to jeopardize access for many to basic schooling, life-long learning, health, and environment safety. Illiteracy also hinders the economic prosperity of the poorest societies in this digital age. Among the underserved population in Latin America, many of the indigenous children are the poorest of the poor who hardly have access to formal and stable schooling. This paper reviews the literature addressing education inequality issues in Latin America and possible opportunities with mobile learning technology to counter the effects of the education inequality. Also, this paper suggests mobile technology design considerations to meet the particular learning needs of the extremely underserved and underachieving indigenous children in Latin America.