Abstract: This patent relates to hinged devices, such as computing devices. One example includes a first portion including a first input/output device and a second portion including a second input/output device. A hinge assembly includes a flexible hinge that removably couples the first and second portions and allows relative rotation between the first and second portions. The flexible hinge is biased into the first portion to reduce a percentage of the flexible hinge exposed between the first and second portions at a given rotational or angular orientation of the first and second portions.

Abstract: A heat exchanger comprising a heatsink and/or coldplate is disposed on a semiconductor having a heat-producing die within. A layer of thermal interface material (TIM) is disposed between the heat exchanger and semiconductor to enhance heat dissipation as the semiconductor is operated. A seal including a gasket or edgebond adhesive is provided around the perimeter edges of the heat exchanger and semiconductor to seal the gap around the periphery of the TIM layer to prevent the TIM from getting pumped out with cyclical thermal loading of the assembly. A capillary tube in the heat exchanger extending from the internal TIM layer to an opening exposed to the surrounding environment provides a reservoir to capture TIM that would otherwise be pumped out. Dimensions of the capillary tube are selected to prevent environmental air from passing by the TIM in the tube and getting entrapped in the TIM layer as voids.

Abstract: A technique accelerates the generative production of tokens using a target language model that operates in cooperation with a draft language model. The target language model is more capable, but slower, compared to the draft language model. In operation, the draft language model transforms prompt tokens into draft tokens. The target language model edits the draft tokens, e.g., by selecting zero, one, or more of the draft tokens, and by also predicting a next token to follow the draft token(s) (if any) that are selected. Further, the target language model produces guidance vector information. In a subsequent cycle, the draft language model uses the guidance vector information to produce an updated set of set of draft tokens. The guidance vector information informs the draft language model of the embedding space being used by the target language model. This achieves a more effective cooperative relation between the two models.

Abstract: Aspects of the present disclosure relate to multi-user, multi-device gaze tracking. In examples, a system includes at least one processor, and memory storing instructions that, when executed by the at least one processor, causes the system to perform a set of operations. The set of operations include identifying a plurality of computing devices, and identifying one or more users. The set of operations may further include receiving gaze input data and load data, from two or more of the plurality of computing devices. The set of operations may further include performing load balancing between the plurality of devices, wherein the load balancing comprises assigning one or more tasks from a first of the plurality of computing devices to a second of the plurality of computing devices based upon the gaze input data.

Abstract: A remote monitoring and management (RMM) system is configured to receive a stream of events generated in response to interactions of users from multiple tenants with one or more applications and store the events in a database. A plurality of different insight types is defined for one or more event types for the events. Insights of the different insight types are generated based on the events in the database, the event types of the events, and numbers of events of the event types. The insights are ranked using an artificial intelligence (AI) model trained to generate a predicted success score for each of the insights. A predetermined number of top insights are selected based on the ranking of the insights and aggregated into a feed. The feed is to at least one computing device associated with the RMM system.

Abstract: This description relates to removing CO2 from the air. One example includes a duct extending from an external environment to an internal environment and a fan configured to move air through the duct. The example also includes first and second CO2 removal assemblies configured to alternatively transition between CO2 adsorption mode and CO2 desorption mode so that one of either the first and second CO2 removal assemblies is in CO2 adsorption mode and receiving at least some of the air moving through the duct while the other of the first and second CO2 removal assemblies is not receiving air moving through the duct while CO2 is removed in desorption mode.

Abstract: Endpoint security groups include computing device endpoints that are classified according to commonly shared device features and capabilities including device type, function, role, or location. Endpoint security groups are used as an alternative identity mechanism for endpoints for purposes of security and data traffic policy enforcement rather than using conventional IP (Internet Protocol) addressing. Grouping endpoints reduces the scope of network management to enable dynamic policy enforcement for endpoints as they join, leave, and then rejoin computing networks, which is a common behavior, particularly for IoT (Internet-of-Things) devices in manufacturing environments. In an illustrative example, a private multi-access edge compute (MEC) platform supports a scalable policy definition and enforcement framework that provides consistent endpoint handling independent of network access methodology.

Abstract: A method for network traffic arbitration includes, at a network router, receiving two or more network packets over two or more input ports. During an observation window, traffic parameters for the two or more network packets are stored in a traffic history table, the traffic parameters including a Quality-of-Service (QOS) priority value for a network packet of the two or more network packets. Based at least in part on the traffic parameters recorded in the traffic history table, including the QoS priority value, arbitration weights are calculated for each of the two or more input ports for a weighted round robin arbitration process.

Abstract: Liquid-cooled coldplates are mounted to racks receiving solid state drives (SSDs) in an electronic component rack. The SSDs have heat spreaders with externally exposed surfaces that are thermally coupled to the coldplates using dry-contact interfaces. The SSD heat spreaders and rack-mounted coldplates provide a thermal path from the heat-producing semiconductors inside the SSD to a fluid distribution system in the rack that is operatively coupled to a liquid-cooling system. The SSDs are slideably mounted in the racks to support easy “hot-swapping.” A technician slides an SSD into the rack racks and uses a finger-operated mechanism in the SSD to simultaneously seat SSD power and data connectors to mating connectors in the rack and place the coldplate in intimate thermal contact with the SSD heat spreader.

Abstract: Endpoint security groups include computing device endpoints that are classified according to commonly shared device features and capabilities including device type, function, role, or location. Endpoint security groups are used as an alternative identity mechanism for endpoints for purposes of security and data traffic policy enforcement rather than using conventional IP (Internet Protocol) addressing. Grouping endpoints reduces the scope of network management to enable dynamic policy enforcement for endpoints as they join, leave, and then rejoin computing networks, which is a common behavior, particularly for IoT (Internet-of-Things) devices in manufacturing environments. In an illustrative example, a private multi-access edge compute (MEC) platform supports a scalable policy definition and enforcement framework that provides consistent endpoint handling independent of network access methodology.

Abstract: A data processing system implements receiving, via a first software application on a client device, a call requesting a schedule to be generated for a user by a generative model. The system further implements identifying online and/or offline data source(s) indicating activities specific to the user, the online and/or offline data source(s) including software application(s) within a workspace; constructing a first prompt by a prompt construction unit as an input to the generative model, the prompt construction unit constructing the first prompt by appending the activities and context data to an instruction string, the instruction string comprising instructions to the generative model to schedule the activities based on the context data, and to assign the scheduled activities into the schedule, the context data being associated with the user and/or the activities; providing the schedule to the client device; and causing a user interface of the client device to present the schedule.

Abstract: A data processing system implements receiving an image and a natural language prompt input by a user requesting that an application generate an digital picture frame for the image; analyzing the prompt using a key-phrase extraction unit to extract one or more key phrases from the prompt that describe a topic of the frame to be generated for the image; providing the one or more key phrases as an input to a retrieval engine; analyzing the one or more key phrases with the retrieval engine to identify a set of candidate frame images from among a plurality of frame images in a labeled frame images datastore; analyzing the set of candidate frame images using an image placement unit to obtain a set of framed images based on the image and the candidate frame images; and presenting the set of framed images on a user interface of the application.

Abstract: A database management system for managing a database includes each document being stored as a number of replicas for accessibility and data preservation. The system includes: a processor; a network interface; and a memory comprising programming instructions for execution by the processor to implement a database management service, the service configured to maintain a primary replica of a document, a number of secondary replicas of the document, and another log-only replica storing a log of changes to the document rather than contents of the document. The service makes head reads to the primary replica as needed when a read request to the number of secondary replicas does not result in a quorum.

Abstract: The disclosed technology is generally directed to a distributed query-and-command system. In one example of the technology, in a trusted execution environment (TEE) of a first node, database code of the first node and distributed ledger code of the first node is executed, such that execution of the distributed ledger code of the first node instantiates a first instance of a distributed ledger of a consortium blockchain, and such that execution of the query-and-command code of the first node instantiates a first instance of a query-and-command system. The consortium blockchain is distributed among a plurality of nodes, and the query-and-command system is distributed among the plurality of nodes. A first transaction that is associated with modifying the query-and-command system is received. The first transaction is executed. Changes associated with the first transaction to the distributed ledger are persisted.

Abstract: A computer-implemented approach for assessing and managing risk of a cloud service is disclosed. Cloud computing resource data for a cloud service is received. A risk assessment framework is applied to the cloud computing resource data. The risk assessment framework includes a set of security criteria including a subset of data plane criteria and a subset of control plane criteria. The risk assessment framework assigns an individual risk score to each security criteria of the set. The individual risk scores of the set of security criteria are aggregated to generate an overall risk score for the cloud service. A graphical user interface including the overall risk score is visually presented via a display. A computer-automated risk management operation that automatically adjusts security settings of the cloud service based at least on the cloud computing resource data for the cloud service is executed to enhance security of the cloud service.

Abstract: Natural language generators (NLGs), including large language models, are powerful technologies that are in widespread use. However, typically, as NLGs become more powerful and sophisticated, their correspondingly increased complexity requires substantial processing resources. The present disclosure provides automated techniques for dynamically routing queries between at least two NLGs based on an assessment of query difficulty. Less difficult queries can be routed to a less resource intensive NLG, while more difficult queries are routed to a more sophisticated, but more resource intensive NLG. Routing less difficult queries to a less resource intensive model can thus conserve computing resources, while providing little to no drop in response quality, and in some cases providing improved response quality.

Abstract: A user can select a capacity setting for a transitional partition that determines the allocation between a low-density partition and a high-density partition in the transitional partition. The transitional partition can dynamically change among multiple settings having different capacities for the low-density partition. If the current setting of the transitional partition does not efficiently utilize the available storage space based on the user's preferences for storing different types of data in the low-density partition and the high-density partition, then the user can choose to change the transitional partition to a different setting that better suits the individual user's storage allocation preferences. Therefore, valuable storage space will not be under-utilized but instead will be repurposed for more efficient use by converting a low-density partition to a high-density partition, and vice versa.

Abstract: Examples of the present disclosure describe systems and methods for preventing illicit data transfer and storage. In aspects, a computing platform may receive a data request from a caller system, device, or service. The computing platform may identify data items/properties associated with the data request and retrieve one or more rules relevant to the caller and/or caller location. The retrieved rule(s) may be used to evaluate the data item(s) such that data items, data item content, and/or data item properties that are prohibited by the retrieved rule(s) from being manipulated (e.g., accessed, transferred, stored) are removed from the identified data item(s). Based on the evaluation of the identified data item(s), one or more relevant status codes may be set. The computing platform may then manipulate the identified data item(s) in accordance with the data request and provide a processing response to the caller.

Abstract: The described technology provides a method including selecting a physical SFT entry to be victimized, the physical SFT entry being one of a base SFT entry and an extra SFT entry of a logical SFT entry in a snoop filter (SFT) and consolidating the base SFT entry and the extra SFT entry into one physical entry.

Abstract: A method for predicting the pose of an articulated object includes receiving spatial information for n joints of the articulated object. The spatial information for the n joints is passed to a machine learning model previously trained to receive spatial information for n+m joints as input, wherein m>=1. From the machine learning model, a pose prediction for the articulated object is received as output based at least on the spatial information for the n joints, and without spatial information for the m joints.