Abstract: A method includes obtaining table data and determining a plurality of sorted tuple indexes based on the table data. Each respective sorted tuple index includes a plurality of key values. The method includes partitioning each respective sorted tuple index of the plurality of sorted tuple indexes into two or more partitions. Each respective partition includes at least one key value from each of the plurality of sorted tuple indexes. For each respective partition of the two or more partitions, the method includes assigning a corresponding worker node to the respective partition and generating, using the corresponding worker node, a corresponding b-tree sub-index based on the respective partition. The method includes generating, using a leader node, a b-tree index based on the corresponding b-tree sub-index generated for each respective partition of the two or more partitions.

Abstract: Techniques and devices for a smart-home management runtime by a hub in a home network are described in which the hub discovers if other hubs are present on the home network. If one or more other hubs are discovered, the hub connects with the one or more other hubs, partitions smart-home hub functions between the hub and the one or more other hubs, and based on the partitioning, provides at least some of the smart-home functions to devices on the home network. If no other hubs are present on the home network, the hub provides the smart-home functions to devices on the home network.

Abstract: A method includes receiving a content feed that includes audio data corresponding to speech utterances and processing the content feed to generate a semantically-rich, structured document. The structured document includes a transcription of the speech utterances and includes a plurality of words each aligned with a corresponding audio segment of the audio data that indicates a time when the word was recognized in the audio data. During playback of the content feed, the method also includes receiving a query from a user requesting information contained in the content feed and processing, by a large language model, the query and the structured document to generate a response to the query. The response conveys the requested information contained in the content feed. The method also includes providing, for output from a user device associated with the user, the response to the query.

Abstract: The present document describes systems and methods for on-device person-recognition and provision of intelligent alerts. The system includes a decentralized multi-camera system for on-device facial recognition. A device (e.g., security camera, video doorbell) captures images/video of a person, processes input image frames, detects face images, filters static faces, and aligns a rotation of the face to be upright and frontal. The device then filters low-quality face images and/or images having a large portion of the face occluded. The device computes a face embedding, compares it against a set of locally stored reference embeddings, and sends matching results to cloud services, which, based on the matching result, notifies the device owner whether the observed person is a known person or a stranger. Face detection and recognition computations are performed on-device, not at the cloud. No sensitive information is transmitted off device and privacy is thus preserved.

Abstract: Systems and methods for continuous insects sensing are described. One example method includes receiving a flow at a singulator, the flow comprising one or more insects; singulating the insects into a single-file flow of insects; sensing, using a sensor, insects within the single-file flow of insects; and incrementing a counter based on each sensed insects in the single-file flow of insects.

Abstract: An acoustic structure for an electronic device routes audio from an internal audio driver to an external environment. The acoustic structure includes an acoustic channel fluidly coupling the internal audio driver to an acoustic outlet located at a periphery of a housing. The acoustic channel includes an inlet adjacent to the audio driver having a triangular shape and extends at an oblique angle toward the acoustic outlet to route acoustic waves around an adjacent internal space. The acoustic channel comprises asymmetric lateral boundaries, featuring a substantially straight first wall and a spaced apart second wall having a curved profile. The curved profile is characterized by alternating positive and negative curvatures. This geometric configuration facilitates the fluid dynamic routing of acoustic waves from an offset internal position, directing the acoustic output to manage sound intensity distribution and acoustic energy transmission.

Abstract: A voltage regulator assembly for an integrated circuit package includes a heatsink positioned above an upper surface of an integrated circuit package. The heatsink defines a cavity above the integrated circuit package and a voltage regulator is positioned inside the cavity defined by the heatsink. A pogo pin pad is positioned between the voltage regulator and integrated circuit package and is configured to transfer a voltage from the voltage regulator to the integrated circuit package. Also provided is a method of regulating voltage for an integrated circuit package using such a voltage regulator assembly.

Abstract: Systems and techniques are disclosed for generating entries for a searchable index based on rules generated by one or more machine-learned models. The index entries can include one or more tokens correlated with an outcome and an outcome probability. A subset of tokens can be identified based on the characteristics of an event. The index may be searched for outcomes and their respective probabilities that correspond to tokens that are similar to or match the subset of tokens based on the event.

Abstract: A method for distributing encryption keys includes receiving a table associated with a particular user, the table including a plurality of data blocks and splitting the table into a plurality of tablets including a corresponding portion of data blocks. The method also includes generating a resource key uniquely associated with the table and for each tablet generating a unique data encryption key for the corresponding tablet to encrypt with the unique data encryption key. The method also includes encrypting each data encryption key with the resource key and distributing control of each encrypted tablet and each corresponding encrypted data encryption key to a plurality of tablet servers, each controlling one or more of the encrypted tablets. The resource key transmits to a remote entity causing the remote entity to encrypt the resource key with a user key associated with the particular user and transmit the encrypted resource key.

Abstract: This document describes techniques and apparatuses for computational photography under low-light conditions for an image-capture device on a mobile computing device. In aspects, described are techniques and apparatuses for an image-capture device to utilize sensor data in determining whether to enable flash photography or capture multiple images of the scene without use of a flash under low-light conditions. In other aspects, an image-capture device may utilize device data in determining whether to enable flash photography or capture multiple images of the scene without use of a flash under low-light conditions. The disclosed techniques and apparatuses may provide improved computational photography under low-light conditions for an image-capture device on a mobile computing device.

Abstract: Actuator assemblies, and electronic devices and wearable devices including actuator assemblies, are disclosed. An actuator assembly includes: a first housing enclosing an audio actuator and a first open volume; and a second housing enclosing a haptic actuator and a second open volume. The first open volume is fluidly connected with the second open volume by an aperture between the first housing and the second housing. A first wall of the first housing abuts a second wall of the second housing. The aperture includes a first opening in the first wall of the first housing and a second opening in the second wall of the second housing. The first housing includes a linking component configured to mechanically connect to the second housing. The linking component includes a fluid conduit extending from the first open volume to an opening in the second wall of the second housing.

Abstract: A computer-implemented method for determining traffic conditions includes receiving, by one or more computing devices, one or more requests for traffic condition information from a user, the one or more requests including data indicative of a first location, a second location, and a traffic type; determining, by the one or more computing devices, a first transit route from the first location to the second location, the first transit route including one or more transit areas associated with a first traffic type from a plurality of traffic types; determining, by the one or more computing devices, map data indicative of a first transit time for the first transit route, the first transit time corresponding to a traffic speed associated with the first traffic type; and providing, by the one or more computing devices, the map data to the user in response to the request for traffic condition information.

Abstract: A method includes executing an initial instance of a change log process for a distributed system, each instance of the change log process configured to store a transaction history of transactions executed on the distributed system. The method also includes receiving transaction requests for executing corresponding transactions on the distributed system and determining a change log load based on the received transaction requests. The method includes executing at least one subsequent instance of the change log process when the change log load satisfied a threshold load. When multiple instances of the change log process are executing, the method includes ceasing execution of the at least one subsequent instance of the change log process and merging the transaction history of the initial instance of the change log process and the transaction history of the at least one subsequent instance of the change log process.

Abstract: Methods, systems and apparatus for estimating the fidelity of a quantum computing system. In one aspect, a method includes defining one or more random quantum circuits, wherein a noisy experimental implementation of each random quantum circuit is approximated by a depolarizing channel with respective polarization parameter; generating, for each defined random quantum circuit and by the quantum computing system, a set of experimental data, wherein data items in the set of experimental data comprise measured bit strings corresponding to experimental implementations of the random quantum circuit; determining, for each of the one or more random quantum circuits, an estimate of the respective polarization parameter, comprising maximizing a log-likelihood of the polarization parameter conditioned on the respective set of experimental data using series inversion; and determining an estimate of the fidelity of the quantum computing system based on the determined estimates of respective polarization parameters.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for combating false advertising using crowdsourcing. In one aspect, a method includes receiving a false digital component alert indicating that a digital component presented at a client device includes false information, validating the false digital component alert based on a public key to verify digital signature included in the false digital component alert matching the public key of stored attestation tokens specifying presentation and interaction data for the digital component. In response, adding the false digital component alert to an aggregated report. Based on a false alert ratio, determining that a magnitude of validated false digital component alerts in the aggregated report meets a threshold, and triggering a false digital component mitigation response including providing a false information warning with the false digital component to one or more client devices or digital component provider.

Abstract: A system and method for power pooling, the system including a shared power line, a plurality of backup power supplies connected to the shared power line through respective transformers and one or more controllers configured to, for each respective backup power supply, after passage of a different respective predetermined amount of time, determine whether a transformer connected to the backup power supply is energized. In response to the transformer being energized, the controllers synchronize the backup power supply with the transformer and then set a switch positioned between the backup power supply and the transformer to a fully closed state. In response to the transformer not being energized, the controllers transition the switch from a fully open state to the fully closed state according to an inrush current limiting scheme.

Abstract: Implementations described herein are directed to leveraging odor sensor(s) of client device(s) in responding to user request(s) and/or in generating notification(s). Processor(s) of a given client device can receive a request to identify an odor in an environment of the given client device, process an odor data instance generated by the odor sensor(s) of the given client device, identify the odor based on processing the odor data instance, generate a response that identifies the odor and/or a source of the odor, and cause the response to the request to be rendered via the given client device. Processor(s) of the given client device can additionally, or alternatively, establish baseline odor(s) in the environment and generate a notification when an odor is detected that does not correspond to the baseline odor(s) and/or exclude the baseline odor(s) in generating the response to the request.

Abstract: A method includes receiving an automated speech recognition (ASR) request from a user device that includes a speech input captured by the user device and content metadata associated with the speech input. The content metadata is generated by the user device. The method also includes determining a priority score for the ASR request based on the content metadata associated with the speech input and caching the ASR request in a pre-processing backlog of pending ASR requests each having a corresponding priority score. The pending ASR requests in the pre-processing backlog are ranked in order of the priority scores. The method also includes providing, from the pre-processing backlog, one or more of the pending ASR requests to a backend-side ASR module, wherein pending ASR requests associated with higher priority scores are processed before pending ASR requests associated with lower priority scores.

Abstract: Described herein is a method for managing media item playback. A method includes presenting a media player playing a media item in a first portion of a user interface (UI) provided by a first application on a screen of a user device, responsive to a first user input, reducing a size of the media player playing the media item to allow a user to open a second mobile application, causing playback of the media item to be continued in the media player of the reduced size while content associated with the second application is being presented to the user on the screen of the user device, and responsive to a second user input, restoring the size of the media player while the media player continues to play the media item.

Abstract: Responsive to receiving an input query at a client device, an execution scheme to comply with a deadline for generating a comprehensive response which is responsive to the input query can be generated. The execution scheme can include a plurality of sub-queries decomposed from the input query, each associated with an estimated execution time. As such, generating the execution scheme can include generating the execution scheme such that the total of the estimated execution times complies with the deadline. The sub-queries can then be processed, based on the execution scheme, to generate corresponding sub-query responses. In response to determining that each of the sub-queries of the execution scheme have been processed, a comprehensive response to the input query can be generated using the sub-query responses. The comprehensive response can then be rendered, at the client device or another client device determined to be associated with the client device.

Abstract: The technology provides a rich attention mechanism for structured information extraction of web pages and other electronic documents. An input layer of a model obtains system, information associated with the document, including field tokens representing respective fields to be extracted from the document, structured document type tokens associated, and text tokens from a text sequence in the document. An encoder connects the field tokens, the S type tokens and the text tokens according to a set of different attention patterns. The encoder generates an overall token representation based on the set of different attention patterns. An output layer of the model extracts a final text span for the each of the respective fields from the set of text tokens. The extracted final text span for each of the respective fields is stored in memory, and can be produced in response to a search query, analytics evaluation or other request.

Abstract: The present disclosure provides a system and method using a charge collection antenna in a wearable device to collect charge variation based on a user's jawbone and muscle motion. The collected charge variation may be used to determine an on-body status of the wearable device. For example, in wireless earbuds, information acquired from a charge collection antenna may be used to determine whether the earbud is worn in-ear by the user. The collected charge variation may also be used to detect jaw motion by the user.

Abstract: A method for presenting supplemental information corresponding to on-demand media content may include receiving an indication from a web browser extension executing within a web browsing application that a media content item is being presented on a content provider website; identifying the media content item being presented on the content provider website; determining whether supplemental content associated with the broadcasted media content item is available; in response to determining that the supplemental content is available, causing the web browser extension to present an indicator that the supplemental content is available; obtaining the supplemental content corresponding to the broadcasted media content item, wherein each piece of the supplemental content is associated with timing information of the broadcasted media content item; and causing a portion of the obtained supplemental content to be presented concurrently with the media content item in response to detecting that the indicator has been selec

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for using cryptography, secure MPC, and a TEE to perform computations in ways that preserve data privacy and protect the security of data of each party that is involved in the computation process. In one aspect, a first MPC system of a cluster of MPC systems receives, from a computing system, a first secret share of input data. The first MPC system sends, to an application running in a TEE separate from the cluster of MPC systems, first secret shares of a set of data items generated based at least in part on the first secret share of the input data. The application is configured to generate secret shares of application output data. The first MPC system performs, in collaboration with the one or more second MPC systems of the cluster, one or more secure multi-party computations.

Abstract: A device may detect an object in a media content item that is displayed on a display device. A device may retrieve search results for the object. A device may generate a ranked list by ranking the search results according to a plurality of signals. A device may initiate display of the ranked list on a user interface of the display device. In response to selection of a search result from the ranked list, a device may pause playback of the media content item and render a webpage that corresponds to the selected search result.

Abstract: While a mobile device is not connected to a network, a media application creates a commit that describes changes to be made to entities. The media application determines identifiers associated with the entities. The media application determines whether the entities are resolved or unresolved. The media application transforms the entities that are resolved by applying the changes described in the commit, where a corresponding entity that is unresolved is not transformed until a preceding commit is applied. In response to the mobile device establishing a network connection, the media application transmits the commit to a server for acceptance. Responsive to receiving a notification that the server successfully applied the commit, the media application records an expired version of the commit. Responsive to determining that the server failed to apply the commit, the media application discards the commit.

Abstract: An image capture device includes an image sensor configured to capture raw image data, an image signal processor configured to process the raw image data to form a plurality of image frames, and one or more processors to implement a microvideo generation system. The microvideo generation system is configured to perform a trimming operation relating to the plurality of image frames according to one or more image criteria, and generate, based on the trimming operation, a microvideo including a still image and a video in a same data file.

Abstract: Techniques are described herein for warm word arbitration between automated assistant devices. A method includes: determining that warm word arbitration is to be initiated between a first assistant device and one or more additional assistant devices, including a second assistant device; broadcasting, by the first assistant device, to the one or more additional assistant devices, an active set of warm words for the first assistant device; for each of the one or more additional assistant devices, receiving, from the additional assistant device, an active set of warm words for the additional assistant device; identifying a matching warm word included in the active set of warm words for the first assistant device and included in the active set of warm words for the second assistant device; and enabling or disabling detection of the matching warm word by the first assistant device, in response to identifying the matching warm word.

Abstract: A color display that can deactivate colors based on one or more criteria is disclosed. Reducing the number of color channels used to display an image may help conserve power, which may help extend the operating life of a battery-operated device. The display includes a backplane that is segmented by color to enable the deactivation. The segmentation may also reduce electrical loading on the electrical lines used to address the pixels in the backplane.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for automated image generation. Methods can include generating, by an artificial intelligence (“AI”) system, image prompts based on (i) a category and (ii) a set of constraints. At least one of the image prompts is removed based on a set of excluded terms that are not allowed to be included in qualified image prompts. A set of images is generated using the set of qualified image prompts. One or more images that include excluded objects are removed to obtain a set of qualified generated images. The set of qualified generated images are stored with data specifying the category. At least one qualified generated image is combined with other content, resulting in a new digital component. The new digital component is transmitted to a client device in response to a request for content in the category.

Abstract: Permission control and management for messaging application bots is described. A method can include providing a messaging application, on a first computing device associated with a first user, to enable communication between the first user and another user, and detecting, at the messaging application, a user request. The method can also include programmatically determining that an action in response to the user request requires access to data associated with the first user, and causing a permission interface to be rendered in the messaging application, the permission interface enabling the first user to approve or prohibit access to the data associated with the first user. The method can include accessing the data associated with the first user and performing the action in response to the user request, upon receiving user input from the first user indicating approval of the access to the data associated with the first user.

Abstract: A method includes receiving a training example for a listen-attend-spell (LAS) decoder of a two-pass streaming neural network model and determining whether the training example corresponds to a supervised audio-text pair or an unpaired text sequence. When the training example corresponds to an unpaired text sequence, the method also includes determining a cross entropy loss based on a log probability associated with a context vector of the training example. The method also includes updating the LAS decoder and the context vector based on the determined cross entropy loss.

Abstract: A computing system obtains a plurality of images and associated image information. Each of the plurality of images depicts a particular Point of Interest (POI). For each of the plurality of images, the image information is descriptive of a plurality of image characteristics, such as a geolocation characteristic indicative of a geolocation at which the image was captured and/or an interaction metric derived from previous user interactions with the image. The computing system selects a particular image of the plurality of images based on the image information and generates a movement pattern for a camera viewpoint within a three-dimensional environment. The movement pattern for the camera viewpoint moves around a three-dimensional representation of the particular POI within the three-dimensional environment, and a starting position for the movement pattern is based, at least in part, on the geolocation at which the particular image was captured.

Abstract: Systems and methods for content snippet generation, storage, and suggestion can include obtaining a user input, segmenting a sub-portion of displayed content based on the user input, and generating a content snippet that includes the segmented content and source data associated with the displayed content. A generative model can be leveraged to categorize the content snippet, determine similar content, and determine when to surface the content snippet to a user.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for fine-tuning a computer vision neural network. In particular, the neural network is fine-tuned using task rewards and through reinforcement learning.

Abstract: A computing unit is disclosed, comprising a first memory bank for storing input activations and a second memory bank for storing parameters used in performing computations. The computing unit includes at least one cell comprising at least one multiply accumulate (“MAC”) operator that receives parameters from the second memory bank and performs computations. The computing unit further includes a first traversal unit that provides a control signal to the first memory bank to cause an input activation to be provided to a data bus accessible by the MAC operator. The computing unit performs one or more computations associated with at least one element of a data array, the one or more computations being performed by the MAC operator and comprising, in part, a multiply operation of the input activation received from the data bus and a parameter received from the second memory bank.

Abstract: A DU and/or a CU of a distributed base station implement the techniques of this disclosure to support early data transmission from a UE operating in an inactive or idle state. In some embodiments, the DU receives (304) a MAC PDU from the UE when the radio connection between the UE and the DU is not active (302). The DU provides (308) the data packet included in the MAC PDU to the CU while preventing (306) the CU from initiating a procedure for resuming the radio connection. In this manner, the base station avoids an unnecessary transition of the UE to the connected state. The CU may then proceed with releasing (312) radio resources or otherwise determining to reactivate the radio connection.

Abstract: Techniques and devices for sharing intelligence-derived information by a hub in a home network are described in which the hub exposes a virtual device including one or more clusters on the home network and receives from a partner device a request to subscribe to a cluster of the one or more clusters. The hub receives state information from an intelligence service, stores the received state information as an attribute of the cluster, and publishes the attribute of the cluster to the partner device, the publishing being effective to direct the partner device to determine whether to perform a local action based on the attribute. Alternatively, the hub receives, from a partner device, an advertisement of a custom cluster installed at the partner device and provides state information by sending a command to the custom cluster at the partner device, the command being indicative of the received state information.

Abstract: This document describes systems and techniques directed at a large language model (LLM) augmented user emergency interface. A user device receives, via a user emergency interface, a free-form input associated with an emergency event. An LLM, at least partially deployed on the user device, extracts one or more relevant details of the emergency event from the free-form input. The free-form input can be a text input, a voice input, or a natural language input. The LLM generates an emergency message based on the one or more relevant details. The user device provides, for an emergency messaging session, the emergency message. The LLM can further be configured to compress the emergency message prior to the providing of the emergency message over a bandwidth-constrained network or a latency-constrained network, such as a satellite communication network.

Abstract: A method includes obtaining proximity information for each of a plurality of assistant-enabled devices within an environment of a user device. Each assistant-enabled device is controllable by an assistant application to perform a respective set of available actions associated with the assistant-enabled device. For each assistant-enabled device, the method also includes determining a proximity score based on the proximity information indicating a proximity estimation of the corresponding assistant-enabled device relative to the user device. The method further includes generating, using the proximity scores determined for the assistant-enabled devices, a ranked list of candidate assistant-enabled devices, and for each corresponding assistant-enabled device in the ranked list, displaying, in a graphical user interface (GUI), a respective set of controls for performing the respective set of actions associated with the corresponding assistant- enabled device. 49 75276250.

Abstract: The technology generally relates to spatial audio communication between devices. For example, a first device and a second device may be connected via a communication link. The first device may capture audio signals in an environment through two or more microphones. The first device may encode the captured audio with direction information. The first device may transmit the encoded audio via the communication link to the second device. The second device may decode the encoded audio to be output by one or more speakers of the second device. The second device may output the decoded audio to recreate positions of the captured audio signals.

Abstract: A method may receive an image representing displayable content for display by an application. A method may execute a layout extraction model using the image as input and generating a list of elements for the image as output, the list of elements including at least a bounding box defining a portion of the image and a role attribute. A method may add the role attribute to a node in an accessibility tree using the list of elements.

Abstract: Techniques are directed to identifying focus points of a body part of a user as alignment features for calibration verification. In some implementations, a body part of the user may be a hand, and the focus points are joints. A calibration verification involves, in some implementations, capturing an image of a user's hand and detecting the joints of the hand. The calibration verification then involves, in such implementations, determining distances between joints and computing a metric based on the distances. This metric is then compared to a ground truth metric and the calibration is evaluated based on the verification.