Abstract: A method includes: receiving data identifying, for each of one or more objects, a respective target location to which a robotic agent interacting with a real-world environment should move the object; causing the robotic agent to move the one or more objects to the one or more target locations by repeatedly performing the following: receiving a current image of a current state of the real-world environment; determining, from the current image, a next sequence of actions to be performed by the robotic agent using a next image prediction neural network that predicts future images based on a current action and an action to be performed by the robotic agent; and directing the robotic agent to perform the next sequence of actions.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating document vector representations. One of the methods includes obtaining a new document; and determining a vector representation for the new document using a trained neural network system, wherein the trained neural network system has been trained to receive an input document and a sequence of words from the input document and to generate a respective word score for each word in a set of words, wherein each of the respective word scores represents a predicted likelihood that the corresponding word follows a last word in the sequence in the input document, and wherein determining the vector representation for the new document using the trained neural network system comprises iteratively providing each of the plurality of sequences of words to the trained neural network system to determine the vector representation for the new document using gradient descent.

Abstract: Systems, methods, devices, and other techniques are described herein for training and using neural networks to encode inputs and to process encoded inputs, e.g., to reconstruct inputs from the encoded inputs. A neural network system can include an encoder neural network, a trusted decoder neural network, and an adversary decoder neural network. The encoder neural network processes a primary neural network input and a key input to generate an encoded representation of the primary neural network input. The trusted decoder neural network processes the encoded representation and the key input to generate a first estimated reconstruction of the primary neural network input. The adversary decoder neural network processes the encoded representation without the key input to generate a second estimated reconstruction of the primary neural network input. The encoder and trusted decoder neural networks can be trained jointly, and these networks trained adversarially to the adversary decoder neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for allocating cache resources according to page-level attribute values. In one implementation, the system includes one or more integrated client devices and a cache. Each client device is configured to generate at least a memory request. Each memory request has a respective physical address and a respective page descriptor of a page to which the physical address belongs. The cache is configured to cache memory requests for each of the one or more integrated client devices. The cache comprises a cache memory having multiple ways. The cache is configured to distinguish different memory requests using page-level attributes of respective page descriptors of the memory requests, and to allocate different portions of the cache memory to different respective memory requests.

Abstract: Aspects of the disclosure are directed to a computation unit implementing a systolic array and configured for detecting errors while processing data on the systolic array. Checksum circuit in communication with a systolic array is configured to compute checksums and perform error detection while the systolic array processes input data. Instead of pre-generating checksums in input matrices, input matrices can be directly fed into the systolic array through the checksum circuit. On the output side, the checksum circuit can generate and compare checksums with checksums in an output matrix generated by the systolic array. Error checking the operations to generate the output matrix can be performed without delaying the operations of the systolic array, and without preprocessing the input matrices.

Abstract: The disclosed technology provides techniques, systems, and apparatus for containing and recovering from uncorrectable memory errors in distributed computing environment. An aspect of the disclosed technology includes a hypervisor or virtual machine manager that receives signaling of an uncorrectable memory error detected by a host machine. The virtual machine manager then uses information received via the signaling to identify virtual memory addresses or memory pages associated with the corrupted memory element so as to allow for containment and recovery from the error.

Abstract: This document describes a thermal-control system that is integrated into a security camera. The thermal-control system includes a combination of heatsinks and thermal interface materials with high thermal conductivities. The thermal-control system may transfer and spread energy from a high thermal-loading condition effectuated upon the security camera to concurrently maintain temperatures of multiple thermal zones on or within the security camera at or below prescribed temperature thresholds.

Abstract: In one embodiment, a system includes first host machines implementing a public-cloud computing environment, wherein at least one of the first host machines includes a resource manager that provides a public-cloud resource interface through which one or more public-cloud clients interact with one or more virtual machines, and second host machines implementing a private-cloud computing environment, wherein at least one of the second host machines includes one or more private-cloud virtual machines, wherein at least one of the first host machines further includes a private-cloud VM resource provider through which the resource manager interacts with the private-cloud virtual machines, wherein the VM resource provider translates requests to perform virtual machine operations from a public-cloud-resource interface to a private-cloud virtual machine interface, and the private-cloud virtual machines perform the requested virtual machine operations in response to receiving the translated requests from the VM resource

Abstract: A method includes receiving a subscription request to obtain notifications of events pertaining to the data file that has been previously uploaded by a user account associated with a user device and stored at the cloud-based file system. The subscription request identifies a destination address for the notifications generated in accordance with the subscription request and one or more events, among multiple events, corresponding to the data file that are to prompt the notifications to be transmitted to the destination address associated with the networked device. The method includes determining that the subscription request is authorized, and identifying an occurrence of an event from the one or more events specified in the subscription request and involving the data file. The method includes transmitting to the destination address a notification indicating the occurrence of the event involving the data file.

Abstract: Techniques of this disclosure are directed to enable a computing device to process voice queries and provide query answers even when the computing device and vehicle do not have internet connectivity. According to the disclosed techniques, a computing device may detect a query via input devices of the computing device and output a query answer determined based on the detected query. Rather than directly querying a remote computing system, various aspects of the techniques of this disclosure may enable the computing device to use a query answer cache to generate the query answer. The query answer cache may include predicted queries and query answers retrieved from a query answer cache of a remote computing system, thereby enabling the computing device to respond to the detected queries while experiencing unreliable internet connection.

Abstract: A method for obtaining server-side and client-side calculations performed in a document includes presenting, on a client computer, the document stored on a server and receiving an input, where the input causes a calculation in the document. The client computer performs the calculation on the client computer and sends the input to the server to concurrently perform the calculation on the server. The client computer obtains a first result from the calculation performed on the client computer and a second result from the calculation performed on the server, selects the first result or the second result as a selected result based on which of the first result or the second result is obtained quicker from the calculation performed on the client computer and the calculation performed on the server, and presents, in the document, the selected result to the user.

Abstract: A method of fabricating a carrier chip for distributing signals among circuit elements of a quantum computing device, includes: providing a multilayer wiring stack, the multilayer wiring stack comprising alternating layers of dielectric material and wiring; bonding a capping layer to the multilayer wiring stack, in which the capping layer includes a single crystal silicon dielectric layer; forming a via hole within the capping layer, in which the via hole extends to a first wiring layer of the multilayer wiring stack; forming an electrically conductive via within the via hole and electrically coupled to the first wiring layer; and forming a circuit element on a surface of the capping layer, in which the circuit element is directly electrically coupled to the electrically conductive via.

Abstract: A method of powering a controller using an intermediate device with power from an environmental system may include receiving current from a power wire from the environmental system; passing the current from the power wire to a second command wire from the controller; monitoring the current flowing between the power wire and the second command wire while the current is below a threshold indicative of an amount of current used to power the controller from the environmental system; detecting when the current flowing between the power wire and the second command wire exceeds the threshold indicating that the controller is sending a command to the environmental system to perform the function; and sending a command to environmental system using a first command wire from the environmental system after detecting that the current exceeds the threshold.

Abstract: Methods, systems and apparatus for identifying advertisements previously presented when processing related search queries related to a submitted search query, and providing the advertisements identified from the related search queries for presentation on a search results page for the submitted search query.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a computer chip placement. One of the methods includes obtaining netlist data for a computer chip; and generating a computer chip placement, comprising placing a respective macro node at each time step in a sequence comprising a plurality of time steps, the placing comprising, for each time step: generating an input representation for the time step; processing the input representation using a node placement neural network having a plurality of network parameters, wherein the node placement neural network is configured to process the input representation in accordance with current values of the network parameters to generate a score distribution over a plurality of positions on the surface of the computer chip; and assigning the macro node to be placed at the time step to a position from the plurality of positions using the score distribution.

Abstract: A circuit for performing neural network computations for a neural network, the circuit comprising: a systolic array comprising a plurality of cells; a weight fetcher unit configured to, for each of the plurality of neural network layers: send, for the neural network layer, a plurality of weight inputs to cells along a first dimension of the systolic array; and a plurality of weight sequencer units, each weight sequencer unit coupled to a distinct cell along the first dimension of the systolic array, the plurality of weight sequencer units configured to, for each of the plurality of neural network layers: shift, for the neural network layer, the plurality of weight inputs to cells along the second dimension of the systolic array over a plurality of clock cycles and where each cell is configured to compute a product of an activation input and a respective weight input using multiplication circuitry.

Abstract: This document describes systems and techniques for encoding a video frame using different compression ratios or compression algorithms for text blocks and non-text blocks. The described systems and techniques can determine, using a machine-learned model, which blocks of a frame include and do not include text. The described systems and techniques can then use a different compression ratio or compression algorithm for text blocks than the compression ratio or compression algorithm used for non-text blocks. For example, the systems and techniques can encode the text blocks using a first compression ratio that results in higher video quality than a second compression ratio used on at least some non-text blocks. In this way, the described systems and techniques can improve text legibility in a video file without significantly increasing the bandwidth requirements to transmit the video file to remote computing devices.

Abstract: A method includes receiving a series of writes that each requests respective data to be stored at a database. For each respective write, the method includes storing the respective data in a first buffer associated with a first data streaming application and a second buffer associated with a second data streaming application. The method includes transmitting the respective data to the database. The method includes receiving, from the database, confirmation that the respective data of the respective write has been committed to the database. The confirmation includes an order indicator indicating an order the respective data of the respective write was committed to the database relative to other writes in the series of writes. In response to receiving the confirmation, the method includes sending the respective data from the first buffer to the first data streaming application and from the second buffer to the second data streaming application.

Abstract: A method for managing cloud infrastructure includes receiving, from a user of a user device, a cloud infrastructure modification request requesting modification to cloud infrastructure. The cloud infrastructure modification request includes abstract configuration data derived from a user interaction with a graphical user interface (GUI) executing on the user device. The method includes translating the abstract configuration data into a configuration command. The configuration command describes a configuration of the cloud infrastructure. The method includes updating a configuration file with the configuration command. The configuration file includes one or more cloud infrastructure specifications for the cloud infrastructure and is controlled by a source control management system. The method includes provisioning, using the updated configuration file, the cloud infrastructure.

Abstract: This document describes improved driving circuits for piezoelectric speakers (piezos). In aspects, a charging case for a hearable device includes a piezo and a circuit for driving the piezo having a battery, an inductor, first and second switches, and a microcontroller. The microcontroller is configured to first toggle the first switch for a number of times to charge the piezoelectric speaker via the inductor to a target voltage level and second toggle the second switch to discharge and thus sound the piezo. In such aspects, the first toggling of the first switch and the second toggling of the second switch control a volume and frequency, respectively, of the sounding, improving audio quality and providing finder features to users. Further, charging the piezoelectric speaker via the inductor enables the inductor to be smaller, conserving space within and power consumption of the charging case, leading to improved battery life.