Adaptive deployment model for hybrid cloud / on-device interaction systems

Abstract

Embodiments are disclosed for an adaptive deployment model for hybrid cloud / on-device systems. In some embodiments, a method comprises: on a device communicatively coupled to a network: monitoring decision parameters related to allocation of artificial intelligence (AI) processing tasks between a network and a device; analyzing, with machine learning, the monitored decision parameters to predict an allocation of AI processing tasks to the network and the device; allocating the AI processing tasks to the network and the device based on the predicted allocation.

Description

Attorney Docket No. 59624-0132WO1ADAPTIVE DEPLOYMENT MODEL FOR HYBRID CLOUD / ON-DEVICE INTERACTION SYSTEMSCROSS-RELATED APPLICATIONS

[0001] This applicant claims the benefit of priority from U.S. Provisional Application No. 63 / 764,533, filed February 27, 2025, for "Adaptive Deployment Model for Hybrid Cloud / On-Device Interaction Systems," which provisional application is incorporated by reference herein in its entirety.TECHNICAL FIELD

[0002] This disclosure relates generally to systems and methods for optimizing computational workload distribution between cloud computing resources and on-device processing capabilities, particularly in applications utilizing artificial intelligence and machine learning models.BACKGROUND

[0003] Hybrid cloud and on-device processing systems offer significant advantages in terms of cost efficiency, latency reduction, and privacy enhancement. However, current implementations typically employ static allocation strategies that fail to adapt to changing operational conditions and requirements.BRIEF DESCRIPTION OF THE DRAWINGS

[0004] FIG. 1A is a block diagram of an adaptive deployment system for hybrid cloud / on-device interactive systems, according to one or more embodiments.

[0005] FIG. IB is a block diagram of alternative adaptive deployment system for hybrid cloud / on-device interactive systems, according to one or more embodiments.

[0006] FIG. 2 is a flow diagram of a process for allocating Al processing tasks between a device and a network, according to one or more embodiments.DETAILED DESCRIPTION

[0007] Embodiments are disclosed for an adaptive deployment model for hybrid cloud / on-device interactive systems. In some embodiments, a method implemented on a device comprises: monitoring decision parameters related to allocation of artificial intelligence (Al) processing tasks between a network and a device; analyzing, with machine learning, theAttorney Docket No. 59624-0132WO1monitored decision parameters to predict an allocation of Al processing tasks to the network and the device; allocating the Al processing tasks to the network and the device based on the predicted allocation.

[0008] In some embodiments, the method further comprises refining at least one decision parameter based on historical performance data related to the device or network and at least one efficiency metric.

[0009] In some embodiments, the method further comprises refining at least one decision parameter based on user behavior parameters or application requirements.

[0010] In some embodiments, the decision parameters include a task complexity requirement.

[0011] In some embodiments, the decision parameters include a network condition.

[0012] In some embodiments, the decision parameters include a latency constraint.

[0013] In some embodiments, the decision parameters include a privacy sensitivity of data being processed.

[0014] In some embodiments, the decision parameters include device resource availability.

[0015] In some embodiments, the decision parameters include a cost implication.

[0016] In some embodiments, the allocating further comprises ensuring data consistency between data stored on the device and data stored on the network.

[0017] In some embodiments, the device is a personal or desktop computer.

[0018] In some embodiments, the device is a computer peripheral device.

[0019] A system implemented on a device comprises: a monitoring unit configured to evaluate one or more decision parameters for allocating artificial intelligence (Al) processing tasks between a device and a network; a decision unit configured to predict, using machine learning, an allocation of the Al processing tasks between the device and the network; and a resource allocation unit configured to implement the allocation of Al processing tasks between the device and the network.

[0020] In some embodiments, the resource allocation unit is further configured to direct or send the Al processing tasks to the device and the network.

[0021] In some embodiments, the resource allocation unit is further configured to implement allocation of Al processing tasks between the device and the network based on user behavior patterns or application requirements.Attorney Docket No. 59624-0132WO1

[0022] In some embodiments, the monitoring unit is further configured to learn from historical performance data of the system and refine the decision parameters based on efficiency metrics.

[0023] In some embodiments, the decision parameters include a task complexity requirement.

[0024] In some embodiments, the decision parameters include a network condition

[0025] In some embodiments, the decision parameters include a latency constraint.

[0026] In some embodiments, the decision parameters include a privacy sensitivity of data being processed.

[0027] In some embodiments, the decision parameters include device resource availability.

[0028] In some embodiments, the device is an Al-enabled personal computer

[0029] Other embodiments include systems, devices, methods and computer-readable mediums.

[0030] Particular embodiments described below provide one or more advantages over existing systems and methods. The adaptive deployment model provides several advantages over static hybrid approaches, including but not limited to: optimized resource utilization based on real-time conditions; reduced operational costs through intelligent workload allocation; enhanced privacy protection by processing sensitive data locally when appropriate; improved user experience through latency optimization; and extended device battery life through contextual resource management

[0031] FIG. 1A is a block diagram of an adaptive deployment system 101 for hybrid cloud / on-device interactive systems, according to one or more embodiments. In some embodiments, system 101 can be embedded in device 100. Device 100 can be any device capable of running all or a portion of an artificial intelligence (Al) model, machine learning model or language model such as a large language model (LLM) or small language model (SLM), including but not limited to: personal computers, tablet computers, smartphones, smart printers, smart speakers or other audio devices, televisions systems, stereo systems, home appliances, loT devices, scanners, cameras, servers, monitors, workstations, thin clients, peripherals (e.g., keyboards, mouse or other I / O devices), audio / visional systems, telepresence and video conference systems, disk drives, docking stations, and wearable devices (e.g., AR / VR glasses / goggles or other eyewear).Attorney Docket No. 59624-0132WO1

[0032] Device 100 is communicatively coupled to network 107. Network 107 includes any local or wide area network (wired, wireless, optical, etc.) and any device connected to or otherwise in communication with the network including server computers, storage devices, hubs, access points, switches, routers, modems, repeaters, firewalls, power management systems, etc.

[0033] In some embodiments, adaptive deployment system 101 includes monitoring unit 102, decision unit 103, resource allocation unit 104 and optimization framework 105.

[0034] Monitoring unit 102 periodically or continuously evaluates various decision parameters locally generated at device 100 and / or by network 107, including but not limited to: task complexity requirements that, for example, describe or specify a complexity such as a computational complexity of an Al or other computing task or workload; network conditions and latency constraints that, for example, describe or specify properties of a network or network connection that will be used by an Al or other computing task or workload; privacy sensitivity of data being processed that, for example, describe of specify privacy properties of an Al or other computing task or workload or data included in or accessed by such a task or workload; device resource availability (e.g., processing power, memory, battery status) that are associated with a device hosting one or more language models or computing systems that access one or more language models to perform or execute an Al or other computing task or workload; and cost implications of cloud resource utilization that, for example, describe or specify financial, environmental, or other costs associated with performing or executing an Al or other computing task or workload by or at a cloud resource such as a computer server or group of computer servers.

[0035] In some embodiments, monitoring unit 102 can determine a changed condition based on the monitored decision parameters. Upon determination of a changed condition, monitoring unit 102 can initiate reallocation of the one or more Al processing tasks to device 100 and / or network 107 (e.g., by alerting decision unit 103), which can result in moving all or some of existing Al processing tasks running on device 100 to network 107 or vice-versa, generating new Al processing tasks which can be allocated to device 100 and / or network 107, terminating all or some existing Al processing tasks running on device 100 and / or network 107, and / or dividing existing Al processing tasks into subtasks and allocating the subtasks to device 100 and / or network 107.

[0036] In some embodiments, reallocation can include predicting, with the machine learning model, a new allocation of Al processing tasks to device 100 and network 107 basedAttorney Docket No. 59624-0132WO1on updated decision parameters, including an increase (new decision parameters) or reduction (decision parameters that are invalid or no longer needed) in the of number decision parameters.

[0037] In some embodiments, monitoring unit 102 monitors resources, for example to determine decision parameters related to those resources, such as network, memory usage and processor usage, and maintains a most recent state of these resources. When a user makes a query, decision unit 103 gets the most recent state of resources, predicts the resources needed for the current query and routes the query based on the parameters. If files or user data referenced is restricted, then decision unit 103 can select a local Al agent based on rules or access control. In some embodiments, decision unit 103 predicts the size of task for each query and allocate long running costly tasks to the local Al agent. Decision unit 103 can also define classes of tasks that are large and need to be done locally by the local Al agent.

[0038] In some embodiments, monitoring unit 102 can keep track of historical performance patterns such as latency and throughput and provide that information into decision unit 103, which can make the best resource allocation not just on the current conditions but on a historical snapshot. The term "learning" as used herein refers to tracking historical parameters and refers to the cumulative parameters when making the decision.

[0039] One or more Al or language models (e.g., LLMs, SLMs, or a combination of LLMs and SLMs) are stored locally (Al models 106) and / or on network 107 (Al models 109). Some examples of Al models 106, 109 include but are not limited to: GPT x.x (e.g., GPT 3.5), AYA, GPT-4o, GPT-4, PHI, Llama or any other suitable machine learning model, framework or technology.

[0040] In some embodiments, one or more local Al models 106 can be replaced with one or more other Al models 109 based on context and need, as described in U.S. Patent Application No. 19 / 048,836. For example, an Al model 106 used by device 100 (e.g., a context-aware device) can be determined from multiple Al models 109 based on context data and / or other data provided by device 100 and / or other sources (e.g., personal information databases). In some embodiments, Al processing tasks are allocated to Al models 109 based on context or needs without replacing Al models 106.

[0041] In some embodiments, Al models 109 can be customized per geographic region to satisfy particular cultures, languages, or service requirements. The interchangeability of one or more Al models 106 with one or more Al models 109 serves the need to properly interact with these cultures, languages, or services. Al models can be determined based on the languageAttorney Docket No. 59624-0132WO1of the user captured by a microphone of device 100 or by other input means (e.g., text analysis or text-to-speech conversion, geographic coordinates).

[0042] In some embodiments, Al models 106, 109 can be determined based on user preferences 110 (see FIG. IB), which can include, for example, a setting selected by the user in a settings menu or other input mechanism. In some embodiments, a language agnostic language model can be used to categorize a request and then forward that request to a particular Al model (e.g., a language model specialized to a particular language) based on the categorization.

[0043] In some embodiments, Al and other data processing tasks can be performed on device 100 and / or server(s) 108, and the allocation of those tasks are determined on device 100. Al processing tasks include but are not limited to any tasks that are performed by device 100 or servers 108 that are related to Al processing, including but not limited to preprocessing of data, training model parameters, and post-processing of data output by Al models 106, 109 (e.g., language models). In other embodiments, non-AI processing tasks are allocated by system 101 between the device and network. System 101 orchestrates the processing between device 100 and server(s) 108 by monitoring the decision parameters that can be used to predict or infer, using machine learning, an allocation of processing tasks between device 100 and server(s) 108.

[0044] In some embodiments, optimization framework 105 learns from historical performance data provided by device 100 and / or server(s) 108 and refines the decision parameters based on, for example, efficiency metrics or other suitable criteria or conditions. Optimization framework 105 refines or adjusts the decision parameters in accordance with user behavior or application requirements. For example, if network 107 is experiencing low bandwidth or instabilities, decision parameters related to network conditions can be adjusted, such that more processing tasks are allocated to device 100.

[0045] In some embodiments, decision unit 103 analyzes monitored decision parameters (e.g., decision parameters monitored by monitoring unity 102) related to Al (e.g., related to Al queries initiated by device 100) using one or more machine learning models 106, predicts or infers an optimal allocation of Al processing tasks between the device and the network, and then initiates allocation of the Al processing tasks in accordance with the predicted allocation as the performance conditions of device 100 and network 107 change. For example, a local Al agent can be configured to do Retrieval-Augmented Generation (RAG) on a list of files and answer user queries based on the knowledge contained in those files. The listAttorney Docket No. 59624-0132WO1of local files (e.g., .pdfs, Word docs, text files, etc.) can be parsed into smaller, manageable chunks, converted into numerical embeddings and stored in a local database (e.g., a vector database).

[0046] In some embodiments, a cloud Al agent available through network 107 can be configured to accept web pages and news sources from the Internet and provide answers to real-time user queries. When the user provides a query, decision unit 103 determines the intent of the user based on the query and determines if the query needs to be answered by the local Al agent and / or the cloud Al agent based on predefined local capabilities 111 (see FIG. IB), which can be stored on the device in the form of a table or other data structure. In this example, predefined local capabilities would exclude international new stories.

[0047] When the user provides a query, decision unit 103 determines if the answer to the query can be formulated by the files in the local database. If the answer can be formulated at least in part from the local database, the local Al agent searches the local database for text most similar to the query. The local Al agent sends the query and the relevant text to local Al model 106. The local Al model 106 uses this text to generate a precise answer to the query. If the answer cannot be formulated at least in part from the local database (e.g., requires knowledge from the Internet), decision unit 103 allocates the processing task to the cloud Al agent to answer the query or the part that cannot be answered by the local Al agent using Al models 109. If both local Al models 106 and remote Al models 109 are used, the answers provided by local Al models 106 and remote Al models 109 are combined by decision unit 103 into a single answer to the user query.

[0048] In some embodiments, one or more outputs of Al models 106 are compared to one or more predetermined thresholds to obtain a decision on allocation of Al processing tasks between device 100 and network 107.

[0049] In some embodiments, resource allocation unit 104 implements workload allocation decisions, manages transitions between cloud and device processing, and ensures data consistency during transitions. This includes allocating (e.g., directing or sending) Al processing tasks to a cloud or remote Al models 109 accessible at servers 108 or to on-device or local Al models 106 accessible at device 100 or adaptive deployment system 101.

[0050] In some embodiments, optimization framework 105 learns from historical performance data of device 100 and network 107, refines decision parameters based on efficiency metrics, and adapts to user behavior patterns and application requirements.Attorney Docket No. 59624-0132WO1

[0051] Adaptive deployment system 101 described above operates by continuously, periodically or response to trigger events or conditions, evaluating the operational context and allocation and reallocating Al processing tasks to maximize efficiency, minimize costs, and maintain application performance requirements.

[0052] FIG. IB is a block diagram of alternative adaptive deployment system for hybrid cloud / on-device interactive systems, according to one or more embodiments. Similar to FIG.1A, the alternative adaptive deployment system 101 includes monitoring unit 102, decision unit 103, resource allocation unit 104 and optimization framework 105. However, in this alternative embodiment, Al models 106 are part of the decision unit 106. Decision unit 103 receives inputs from monitoring unit 102, optimization framework 105, user preferences 110 and predefined local capabilities 111. Resource allocation unit 104 is coupled to monitoring unit 102 and provides resource allocation information to monitoring unit 102.

[0053] Also shown in FIG. IB is local context 112, context manager 113 and cloud context 114. Local context 112 represents the aggregate information that is available locally at device 100 for allocating Al processing tasks between device 100 and network 107. Cloud context 114 represents the aggregate information that is available to network 107 for allocating Al processing tasks between device 100 and network 107. For example, there may be some information in local context 112 that is confidential on the device (e.g., personal user information) that should be processed on device 100 using local Al models 106 rather than network 107 using Al models 109.

[0054] In another example, cloud context 114 may include information regarding the performance or capabilities of network 107 that are not shared with device 100, so such information is not part of local context 112. In some embodiments, certain information may be part of both local context 112 and cloud context 114. In some embodiments, context manager 113 resides on device 100 and understands both local context 112 and cloud context 114 to ensure that the allocation of Al processing tasks is performed quickly and efficiently without violating any confidentiality or security protocols.

[0055] FIG. 2 is a flow diagram of a process for allocating Al processing tasks between a device and a network, according to one or more embodiments. In some embodiments, process 200 is implemented on a device and includes: monitoring decision parameters related to allocation of artificial intelligence (Al) processing tasks between the network and the device (201); analyzing, with machine learning, the monitored decision parameters to predict anAttorney Docket No. 59624-0132WO1allocation of Al processing tasks to the network and the device (202); allocating the Al processing tasks to the network and the device based on the predicted allocation (203).

[0056] In some embodiments, a monitoring unit 102 (See FIGS. 1A, IB) monitors the decision parameters, to determine decision parameters related to those resources, such as network, memory usage and processor usage, and maintains a most recent state of these resources. In some embodiments, monitoring unit 102 can keep track of historical performance patterns such as latency and throughput.

[0057] In some embodiments, the analyzing and allocation prediction is performed by decision unit 103 (See FIGS. 1 A, IB). For example, when a user makes a query, decision unit 103 gets the most recent state of resources, predicts the resources needed for the current query and routes the query based on the decision parameters. If files or user data referenced is restricted, then decision unit 103 can select a local Al agent based on rules or access control. In some embodiments, decision unit 103 predicts the size of task for each query and allocates long running costly tasks to the local Al agent. Decision unit 103 can also define classes of tasks that are large and need to be done locally by a local Al agent.

[0058] In some embodiments, the allocation of Al processing tasks is performed by resource allocation unit 104 (See FIGS. 1A, IB). For example, resource allocation unit 104 implements workload allocation decisions, manages transitions between cloud and device processing, and ensures data consistency during transitions. This includes allocating (e.g., directing or sending) Al processing tasks to a cloud or remote Al models 109 accessible at servers 108 or to on-device or local Al models 106 accessible at device 100 or adaptive deployment system 101.

[0059] Various implementations of the apparatus and methods described herein can be realized in digital electronic circuitry, integrated circuitry, specially designed ASICs, computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0060] These computer programs, also known as programs, software, software applications or code, include machine instructions for a programmable processor, and can be implemented in a high-level procedural and / or object-oriented programming language, and / or inAttorney Docket No. 59624-0132WO1assembly / machine language. As used herein, the terms “machine-readable medium” “computer-readable medium” refers to any computer program product, apparatus and / or device, e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs) used to provide machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0061] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device, e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to the user and a keyboard and a pointing device, e.g., a mouse or a trackball, by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0062] The systems and techniques described herein can be implemented in a computing system that includes a back-end component, e.g., as a data server, or that includes a middleware component such as an application server, or that includes a front-end component such as a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here, or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication such as, a communication network. Examples of communication networks include a local area network (“LAN”), a wide area network (“WAN”), and the Internet.

[0063] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

[0064] Program instructions stored in the memory, along with configuration data may control overall operation of the system. Server computer systems include one or more processing devices (e.g., microprocessors), a network interface and a memory (all not illustrated). Server computer systems may physically take the form of a rack mounted card and may be in communication with one or more operator terminals (not shown).Attorney Docket No. 59624-0132WO1

[0065] All or part of the processes described herein and their various modifications (hereinafter referred to as “the processes”) can be implemented, at least in part, via a computer program product, i.e., a computer program tangibly embodied in one or more tangible, physical hardware storage devices that are computer and / or machine-readable storage devices for execution by, or to control the operation of, data processing apparatus, e.g., a programmable processor, a computer, or multiple computers. A computer program can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a network.

[0066] Actions associated with implementing the processes can be performed by one or more programmable processors executing one or more computer programs to perform the functions of the calibration process. All or part of the processes can be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) and / or an ASIC (applicationspecific integrated circuit).

[0067] Processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer. Generally, a processor will receive instructions and data from a read-only storage area or a random access storage area or both. Elements of a computer (including a server) include one or more processors for executing instructions and one or more storage area devices for storing instructions and data. Generally, a computer will also include, or be operatively coupled to receive data from, or transfer data to, or both, one or more machine-readable storage media, such as mass storage devices for storing data, e.g., magnetic, magneto-optical disks, or optical disks.

[0068] Tangible, physical hardware storage devices that are suitable for embodying computer program instructions and data include all forms of non-volatile storage, including by way of example, semiconductor storage area devices, e.g., EPROM, EEPROM, and flash storage area devices; magnetic disks, e.g., internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks and volatile computer memory, e.g., RAM such as static and dynamic RAM, as well as erasable memory, e.g., flash memory.

[0069] In addition, the logic flows depicted in the figures do not require the particular order shown, or sequential order, to achieve desirable results. In addition, other actions may beAttorney Docket No. 59624-0132WO1provided, or actions may be eliminated, from the described flows, and other components may be added to, or removed from, the described systems. Likewise, actions depicted in the figures may be performed by different entities or consolidated.

[0070] Elements of different embodiments described herein may be combined to form other embodiments not specifically set forth above. Elements may be left out of the processes, computer programs, Web pages, etc. described herein without adversely affecting their operation. Furthermore, various separate elements may be combined into one or more individual elements to perform the functions described herein.

[0071] Other embodiments are within the scope of the following claims. For example, the actions recited in the claims can be performed in a different order and still achieve desirable results. As one example, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In certain implementations, multitasking and parallel processing may be advantageous.

[0072] A system of one or more computers can be configured to perform particular actions by virtue of having software, firmware, hardware, or a combination of them installed on the system that in operation causes or cause the system to perform the actions. One or more computer programs can be configured to perform particular actions by virtue of including instructions that, when executed by data processing apparatus, cause the apparatus to perform the actions.

[0073] While this specification contains many specific implementation details, these should not be construed as limitations on the scope of any inventions or of what may be claimed, but rather as descriptions of features specific to particular embodiments of particular inventions. Certain features that are described in this specification in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable sub combination. Moreover, although features may be described above as acting in certain combinations and even initially claimed as such, one or more features from a claimed combination can in some cases be excised from the combination, and the claimed combination may be directed to a sub combination or variation of a sub combination.

[0074] Similarly, while operations are depicted in the drawings in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may beAttorney Docket No. 59624-0132WO1advantageous. Moreover, the separation of various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products. What is claimed:

Claims

Attorney Docket No. 59624-0132WO1CLAIMS1. A method comprising:on a device communicatively coupled to a network:monitoring decision parameters related to allocation of artificial intelligence (Al) processing tasks between the network and the device;analyzing, with machine learning, the monitored decision parameters to predict an allocation of Al processing tasks to the network and the device; and allocating the Al processing tasks to the network and the device based on the predicted allocation.

2. The method of claim 1, further comprising:refining at least one decision parameter based on historical performance data related to the device or network and at least one efficiency metric.

3. The method of claim 1, further comprising:refining at least one decision parameter based on user behavior parameters or application requirements.

4. The method of claim 1, wherein the decision parameters include a task complexity requirement.

5. The method of claim 1, wherein the decision parameters include a network condition6. The method of claim 1, wherein the decision parameters include a latency constraint.

7. The method of claim 1, wherein the decision parameters include a privacy sensitivity of data being processed.

8. The method of claim 1, wherein the decision parameters include device resource availability.Attorney Docket No. 59624-0132WO19. The method of claim 1, wherein the decision parameters include a cost implication.

10. The method of claim 1, where the allocating further comprises ensuring data consistency between data stored on the device and data stored on the network.

11. The method of claim 1, wherein the device is an artificial intelligence enabled personal computer.

12. The method of claim 1, wherein the device is an audio device.

13. A system comprising:a network;a device communicatively coupled to the network, the device including:a monitoring unit configured to evaluate one or more decision parameters for allocating artificial intelligence (Al) processing tasks between the device and the network;a decision unit configured to predict, using machine learning, an allocation of the Al processing tasks between the device and the network; anda resource allocation unit configured to implement the allocation of Al processing tasks between the device and the network.

14. The system of claim 13, wherein the resource allocation unit is further configured to direct or send the Al processing tasks to the device and the network.

15. The system of claim 13, wherein the resource allocation unit is further configured to implement allocation of Al processing tasks between the device and the network based on user behavior patterns or application requirements.

16. The system of claim 13, wherein the monitoring unit is further configured to learn from historical performance data of the system and refine the decision parameters based on efficiency metrics.Attorney Docket No. 59624-0132WO117. The system of claim 13, wherein the decision parameters include a task complexity requirement.

18. The system of claim 13, wherein the decision parameters include a network condition19. The system of claim 13, wherein the decision parameters include a latency constraint.

20. The system of claim 13, wherein the decision parameters include a privacy sensitivity of data being processed.

21. The system of claim 13, wherein the decision parameters include device resource availability.

22. The system of claim 13, wherein the device is an Al-enabled personal computer.

Images