Group personas generation using machine learning models

Abstract

Aspects of the disclosure arc directed to generating group personas by prompting generative models, such as large generative models. Generating a group persona can include generating a plurality of persona candidates and inter-persona dynamics for the persona candidates based on a description of desired characteristics for the group persona. Various applications can benefit from utilizing the group persona, as opposed to a combined plurality of individual personas, as the group persona can include inter-persona dynamics for improved accuracy in the various applications. Further, processing can be improved, and memory usage reduced, as the generative model may only need to be prompted once to generate the group persona.

Description

GROUP PERSONAS GENERATION USING MACHINE LEARNING MODELSBACKGROUND

[0001] A persona refers to data describing characteristics of an individual, such as how that individual is perceived by other individuals. Personas can also extend beyond simple heuristics to portraying a digital trail of documents and / or media associated with the individual. Generative models can allow for richer and more accessible techniques to generate personas. However, generative models can typically generate individual personas when prompted. The generative models can be prompted numerous times to generate a plurality of individual personas. This generation of multiple individual personas lacks cohesion, as the generative model may not incorporate interpersonal dynamics for the individual personas that reflect how individuals characterized by the individual personas would interact with one another. Further, prompting a generative model numerous times adds significant processing cost and memory usage due to running the generative model numerous times and storing each individual persona generated.BRIEF SUMMARY

[0002] Aspects of the disclosure are directed to generating group personas by prompting generative models, such as large generative models. Generating a group persona can include generating a plurality of persona candidates and inter-persona dynamics for the persona candidates based on a description of desired characteristics for the group persona. Various applications can benefit from utilizing the group persona, as opposed to a combined plurality of individual personas, as the group persona can include inter-persona dynamics for improved accuracy in the various applications. Further, processing can be improved, and memory usage reduced, as the generative model may only need to be prompted once to generate the group persona.

[0003] An aspect of the disclosure provides for a method for generating a group persona including: receiving, by one or more processors, data associated with a description of characteristics for the group persona; generating, by the one or more processors, a plurality of individual personas using a machine learning model based on the data associated with the description of characteristics for the group persona; generating, by the one or more processors, inter-persona dynamics for the plurality of individual personas using the machine learning model based on the data associated with the description of characteristics for the group persona, the inter-persona dynamics andplurality of individual personas forming the group persona; and outputting, by the one or more processors, the group persona.

[0004] Another aspect of the disclosure provides for a system including: one or more processors; and one or more storage devices coupled to the one or more processors and storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations for the method for generating a group persona.

[0005] Yet another aspect of the disclosure provides for a non-transitory computer readable medium for storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations for the method for generating a group persona.

[0006] In an example, the machine learning model includes at least one of a large foundation model, large language model, or large graphical model. In another example, the machine learning model is fine-tuned on at least one of real-world or synthetic data associated with characteristics of various individual and group personas. In yet another example, the data associated with the description of characteristics for the group persona includes at least one of detail for one or more persona styles, images for one or more personas, or dependencies between one or more of the personas. In yet another example, the data associated with the description of characteristics for the group persona comprises more detail about an individual persona than another individual persona.

[0007] In yet another example, the method further includes iteratively refining, by the one or more processors, one or more of the plurality of individual personas through outputting prompts. In yet another example, the method further includes iteratively refining, by the one or more processors, the inter-persona dynamics through at least one of outputting prompts or simulating interactions between the plurality of individual personas.

[0008] In yet another example, the inter-persona dynamics include relationships among the plurality of individual personas. In yet another example, generating the inter-persona dynamics for the plurality of individual personas includes using a graph neural network to model the relationships among the plurality of individual personas, wherein nodes of the graph neural network represent individual personas and edges in the graph neural network represent relationships between the individual personas. In yet another example, the inter-persona dynamics comprise information spread and usage among the plurality of individual personas. In yet another example, generating the inter-persona dynamics for the plurality of individual personas includes using random graph processing to model the information spread and usage.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] FIG. 1 depicts a block diagram of an example group persona generator according to aspects of the disclosure.

[0010] FIG. 2 depicts a block diagram of a group persona generation system according to aspects of the disclosure.

[0011] FIG. 3 depicts a block diagram of an example environment for implementing a group persona generation system according to aspects of the disclosure.

[0012] FIG. 4 depicts a block diagram illustrating one or more machine learning model architectures according to aspects of the disclosure.

[0013] FIG. 5 depicts a flow diagram of an example process for generating a group persona according to aspects of the disclosure.DETAILED DESCRIPTION

[0014] The technology relates generally to using machine learning models to generate group personas. A group persona may refer to data representing multiple individual personas, where each individual persona can coherently interact with, share common characteristics with, and / or share dependencies and / or relationships with one or more other individual personas of the group persona. Generating group personas can improve processing and reduce memory usage in numerous applications. For example, in optimizing end-systems, such content recommendations, the machine learning models can generate a group persona to test out an end-system, rather than individual personas to respectively test out the end-system. The end-system can be run using the group persona, rather than run numerous times per individual persona, to develop a more accurate understanding of the end-system with less processing cost and memory usage. In addition, the dependencies and / or relationships represented by the group persona can more accurately simulate different types of interactions between its component individual personas, as compared with individual personas that are generated independently of one another.

[0015] Generating a group persona can include receiving a description of desired characteristics of the group persona. The description can be received in any manner, such as through a natural language interface, a structured query language interface, and / or a visual input / output interface including one or more sliders, checkboxes, and / or templates. For example, descriptions like“Please generate several characters that simulate the typical hospital environment” or “Create an audience for an upcoming talk on the subject of machine learning” can be provided via a user device through a natural language interface.

[0016] The description can include any level of detail of the desired characteristics. The description can be a single sentence as noted in the examples above. The description can further include desired specifics of one or more persona styles, images for one or more personas, and / or any other detail that may be considered relevant for generation of the group persona. The description can also include various levels of detail for each of the individual personas of the group persona. For example, individual personas with particular traits deemed more important may include more detail in the description than individual personas with traits deemed less important. Individual personas with more detail in the description can become more frequently connected to other individual personas in the group persona, can be featured more prominently in the group persona, and / or be associated with stronger relationships relative to other more weakly connected individual personas in the group persona. As another example, relationships and / or dependencies between individual personas may be included in the description.

[0017] In response to receiving the description, a machine learning model can process the description to output a set of persona candidates to represent the group persona. Example machine learning models can include large language models, large foundation models, large graphical models, and / or any other large generative models. The machine learning model can be fine-tuned on real-world and / or synthetic data associated with characteristics of various individual and group personas. In some examples, the machine learning model is a pipeline of multiple models, trained end-to-end for processing the description to output the set of persona candidates.

[0018] In outputting a set of persona candidates, the machine learning model can initially output a list of personas with general traits, such as name, characteristics, profile, image, etc. The list of personas can include a configurable level of detail for each persona in the list. For example, the machine learning model can output the list of personas as well as corresponding prompts to further refine the persona candidates. Referring back to the example, “Please generate several characters that simulate the typical hospital environment”, the machine learning model can output two doctors, three nurses, and two administrators. The machine learning model can also output prompts next to each doctor, nurse, and administrator for further refinement of these personas, such that further description can be received, such as “Please add that the first doctor is friendly”, “Pleaseadd that the second nurse is new to their job position”, or “Please add that the first administrator did not receive previous training prior to joining the hospital”. The machine learning model can iteratively output the list of personas and prompts in response to further refinement from the prompts until receiving a stopping criterion, such as a notification in response to the prompts that the list is sufficient, a maximum number of iterations has occurred, and / or a maximum amount of time has occurred.

[0019] Once the set of persona candidates is generated, or sufficiently further refined, the machine learning model can develop inter-persona dynamics for the persona candidates through network graphs to form the group persona. For example, the machine learning model can generate prompts based on the list of personas. These relationship prompts can be in addition to or part of the prompts for refining individual personas. Again, referring back to the example, “Please generate several characters that simulate the typical hospital environment”, the machine learning model can output relationship prompts, such that further description on the inter-persona dynamics can be received, such as “Please add that the first doctor and the second doctor are close friends” or “Please add that the first nurse is not friendly with the other two nurses”. As another example, the machine learning model can include a graph neural network that can model the individual personas in a social network, where nodes are the individual personas, and the edges can represent the strength of relationship between individual personas. The machine learning model can model the individual personas in a social network via forest fire, Barabasi- Albert, and / or Erdos-Renyi processing, as examples. Here, the machine learning model can output probabilistic graph candidates to represent the relationships in the group persona.

[0020] The inter-persona dynamics can include how information spreads between the individual personas and how the individual personas use the information to communicate with one another. For example, the machine learning model can also generate prompts based on the list of personas. The information spread prompts can also be in addition to or part of the prompts or refining individual personas. For example, the machine learning model can output the information spread prompts, such as that further description on information spread and usage can be received, such as “Please note the first doctor will not provide patient information to the third nurse” or “Please note the second administrator provides information directly to the doctors rather than to the nurses”. As another example, the machine learning model can utilize a neural network, such as the graph neural network, to output a probability distribution that models the information spread and usage.

[0021] As with the generation of the candidate personas, the machine learning model can iteratively output the inter-persona dynamics and / or the group persona as a whole until receiving a stopping criterion, such as a notification in response to the prompts that the group persona is sufficient, a maximum number of iterations has occurred, and / or a maximum amount of time has occurred. For example, the machine learning model can automatically prompt itself to determine how the individual personas interact in the group persona based on a simulated occurrence. The machine learning model can iteratively prompt itself until hitting a stopping criterion. Simulated interactions can further be manually reviewed, such as through a user interface on a user device, to further refine the group persona. For example, for reviewing an interaction between two individual personas, the user interface can include prompts to indicate whether the interaction was acceptable or not, such as thumbs up / thumbs down icons, check boxes, and / or text boxes to provide comments or further detail on the interaction. In response to one or more stopping criteria being met, the machine learning model can output the group persona.

[0022] The group personas can be utilized in various applications, such as simulating group characteristics in environments such as sales, advertising, employment, politics, teaching, gaming, etc. As an example, a user may want to generate several personas that can debate about a subject, requiring generating personas that share different beliefs about that subject, with some complementing each other but others with opposing views. Without the ability to generate group personas, each individual persona would need to be manually created through trial and error with their interaction estimated at the end. This can result in additional processing cost and memory usage to iteratively run persona generation per individual and subsequently interaction estimation. However, with group persona generation, the several personas can be generated to debate about a subject with less processing cost and memory usage. Further, descriptions for the several personas can include various levels of detail, such as more detail for the personas that participate in the debate and less for the personas that listen to the debate. This group persona can be utilized in preparing a presentation or to simulate audience questions, as examples.

[0023] As another example, a user may want to generate several non-playable characters for a narrative as a part of a video game or table-top game. With group persona generation, a more diverse set of non-playable characters can be generated with varied styles of interaction among themselves as well as with the player character. This can result in a more varied experience with less processing cost and memory usage. As yet another example, a user for an analytics companymay want to simulate a user base for a new app or website. With group persona generation, that user base can be simulated with less processing cost and memory usage. Group persona generation can further simulate how the new app or website is shared among the individual personas in the group, such as to better predict how the new app or website may become viral.

[0024] As yet another example, a group persona for hospital simulation may be used as part of testing an electronic medical records (EMR) system. A group persona can simulate how hospital employees, represented as component individual personas of the hospital group persona, interact with the EMR system, which can vary depending on the characteristics of each individual persona. For example, one nurse may input all of their records during a specific time of day, while another nurse after each patient visit. These and other characteristics can be input as prompts to the machine learning model. These simulated behaviors can determine how the EMR system handles under load or can help determine whether certain future features should be prioritized based on how the group persona interacts with the current system.

[0025] FIG. 1 depicts a block diagram of an example group persona generator 100. The group persona generator can include one or more generative models 102. The generative models 102 can be general usage models and / or can be models fine-tuned to the task of generating a group persona. If fine-tuned, the models 102 can be trained on real-world and / or synthetic data associated with characteristics of various individual and group personas. Example generative models can include large generative models, such as large language models, large foundation models, and / or large graphical models.

[0026] The generative models 102 can receive a prompt 104 to generate a group persona. The prompt 104 can include a description of desired characteristics of the group persona at any level of detail, such as ranging from a single sentence for generating the entire group persona to detailed paragraphs and / or images depicting one or more individual personas that can form the group persona. The prompt 104 can further include a description of desired inter-persona dynamics for the group persona, including how information spreads between individual personas of the group as well as how the individual personas utilize the information.

[0027] In response to the prompt 104, the generative models 102 can output a generated group persona 106. The group persona 106 can include a list of individual personas with the desired characteristics and inter- persona dynamics for the individual personas. The level of detail of the group persona 106 can correspond to the level of detail provided in the prompt 104.

[0028] The generative models 102 can receive one or more additional prompts 108 for refining the generated group persona 106. The additional prompts 108 can include a description for adjusting the characteristics and / or inter-persona dynamics for the individual personas to better reflect the desired characteristics and inter-persona dynamics of the group persona. The description can be at any level of detail, such as ranging from adjusting a single individual persona of the group to adjusting inter-persona dynamics of the group persona overall.

[0029] In response to the one or more additional prompts 108, the generative models 102 can output a refined group persona 106. The generative models 102 can iteratively output refined group personas 106 until one or more stopping criteria are met.

[0030] FIG. 2 depicts a block diagram of a group persona generation system 200. The group persona generation system 200 can be implemented on one or more computing devices in one or more locations.

[0031] The group persona generation system 200 can be configured to receive input data 202. For example, the group persona generation system 200 can receive the input data 202 as part of a call to an application programming interface (API) exposing the group persona generation system 200 to one or more computing devices. The input data 202 can also be provided to the group persona generation system 200 through a storage medium, such as remote storage connected to the one or more computing devices over a network. The input data 202 can further be provided as input through a user interface on a client computing device coupled to the group persona generation system 200. The user interface can include a natural language interface, such as one or more text boxes, and / or a graphical interface, such as one or more sliders, checkboxes, and / or templates.

[0032] The input data 202 can include generation data associated with a description of desired characteristics of a group persona. The desired characteristics of the group persona can be at any level of detail. For example, the desired characteristics of the group persona can be a single sentence to generate a group persona simulating a particular environment, such as a hospital or audience. As another example, the desired characteristics of the group persona can include one or more paragraphs, images, and / or other documentation for generating the group persona, such as specifics about particular persona styles, images depicting particular personas, and / or documents creating a digital trail for particular personas.

[0033] The description of the desired characteristics of the group persona can include desired characteristics for a plurality of individual personas that form the group persona. The desiredcharacteristics of the individual personas can be at any level of detail. Further, the level of detail for each individual persona can vary, where some individual personas may include more detail in the description than other individual personas.

[0034] The description of the desired characteristics of the group persona can include desired interpersona dynamics for the plurality of individual personas. The desired inter-persona dynamics can be at any level of detail. The inter-persona dynamics can include relationships and / or dependencies between one or more of the individual personas. The inter-persona dynamics can further include how information spreads between one or more of the individual personas as well as how one or more individual personas utilize the information, such as in communicating with one another.

[0035] The input data 202 can further include refinement data associated with iteratively adjusting outputted characteristics of the group personas, any of the individual personas, and / or interpersona dynamics. The refinement data can include additional details and / or corrective details for the desired characteristics.

[0036] From the input data 202, the group persona generation system 200 can be configured to output one or more results generated as output data 204. As an example, the group persona generation system 200 can be configured to send the output data 204 for display on a client or user display. As another example, the group persona generation system 200 can be configured to provide the output data 204 as a set of computer-readable instructions, such as one or more computer programs. The computer programs can be written in any type of programming language, and according to any programming paradigm, e.g., declarative, procedural, assembly, object- oriented, data-oriented, functional, or imperative. The computer programs can be written to perform one or more different functions and to operate within a computing environment, e.g., on a physical device, virtual machine, or across multiple devices. The computer programs can also implement functionality described herein, for example, as performed by a system, engine, module, or model. The group persona generation system 200 can further be configured to forward the output data 204 to one or more other devices configured for translating the output data into an executable program written in a computer programming language. The group persona generation system 200 can also be configured to send the output data 204 to a storage device for storage and later retrieval.

[0037] The output data 204 can include candidate individual personas and inter-persona dynamics based on the description of desired characteristics of the group persona. The output data 204 can further include iterations of the individual personas and inter-persona dynamics based on theadditional details and / or corrective details for the group persona. The output data 204 can also include a final outputted group persona, including the plurality of individual personas and inter- persona dynamics with the desired characteristics, once the candidate individual personas and inter-persona dynamics are sufficiently adjusted.

[0038] The group persona generation system 200 can include a candidate personas engine 206, a candidate refinement engine 208, an inter-persona dynamics engine 210, and a group refinement engine 212. The candidate personas engine 206, candidate refinement engine 208, inter-persona dynamics engine 210, and group refinement engine 212 can be implemented as one or more computer programs, specially configured electronic circuitry, or any combination thereof.

[0039] The candidate personas engine 206 can be configured to generate a plurality of individual personas for the group persona. The candidate personas engine 206 can process the generation data associated with the description of desired characteristics to generate a plurality of candidate personas that can form the group persona. The candidate personas engine 206 can process the generation data using a generative model, such as a large generative model. The generative model can correspond to the one or more generative models 102 as depicted in FIG. 1. The plurality of candidate personas can have one or more traits, such as name, attributes, profile, image, and / or documentation forming a digital trail.

[0040] The candidate refinement engine 208 can be configured to adjust the candidate personas to further align the candidate personas with the desired characteristics for the group persona. The candidate refinement engine 208 can output the candidate personas along with prompts for further refining the personas. For example, the prompts can include text boxes, sliders, and / or checkboxes to provide further detail and / or correct aspects of the candidate personas. The candidate refinement engine 208 can adjust the candidate personas by generating further detail, such as additional traits, for one or more of the candidate personas and / or correcting aspects of the candidate personas, such as correcting traits. The candidate refinement engine 208 can adjust the candidate personas using the generative model that generated the candidate personas, using an additional model fine-tuned for refining candidate personas. The candidate refinement engine 208 can iteratively adjust the candidate personas until receiving a stopping criterion, such as a notification that the candidate personas are sufficient, a maximum number of iterations has occurred, and / or a maximum amount of time has occurred.

[0041] The inter-persona dynamics engine 210 can be configured to generate inter-persona dynamics for the group persona. The inter-persona dynamics engine 210 can process the generation data associated with the description of desired characteristics along with the generated candidate personas to generate the inter-persona dynamics. The inter-persona dynamics engine 210 can process the generation data using the generative model, such as the one or more generative models 102 as depicted in FIG. 1. The inter-persona dynamics engine 210 can generate interpersona dynamics through network graphing via a graph neural network that models the candidate personas in a social network, such as via forest fire, Barabasi-Albert, and / or Erdos-Renyi processing. The nodes in the graph can represent the candidate personas and the edges in the graph can represent relationships between the candidate personas. The graph can further be a probabilistic graph representing the relationships and / or dependencies between candidate personas.

[0042] The inter-persona dynamics engine 210 can further generate how information spreads among the candidate personas as well as how the candidate personas utilize the information to interact with one another. The inter-persona dynamics engine 210 can generate the information spread and utilization via prompt generation and / or random graph processing to output a probability distribution that models information spread and usage.

[0043] The group refinement engine 212 can be configured to adjust the inter-persona dynamics to further align the inter-persona dynamics with the desired characteristics for the group persona. The group refinement engine 212 can output the inter-persona dynamics along with prompts for further refining the dynamics. For example, the prompts can include text boxes, sliders, and / or checkboxes to provide further detail and / or correct aspects of the inter-persona dynamics. The group refinement engine 212 can adjust the inter-persona dynamics by generating further detail, such as adding nodes and / or edges to a network graph, or correcting detail, such as switching edges in the network graph. The group refinement engine 212 can adjust the inter-persona dynamics using the generative model that generated the inter-persona dynamics or using an additional model fine-tuned for refining inter-persona dynamics. The group refinement engine 212 can iteratively adjust the inter-persona dynamics until receiving a stopping criterion, such as a notification that the inter-persona dynamics are sufficient, a maximum number of iterations has occurred, and / or a maximum amount of time has occurred. The group refinement engine 212 can then output the group persona.

[0044] FIG. 3 depicts a block diagram of an example environment 300 for implementing a group persona generation system 318. The group persona generation system 318 can be implemented on one or more devices having one or more processors in one or more locations, such as in server computing device 302. Client computing device 304 and the server computing device 302 can be communicatively coupled to one or more storage devices 306 over a network 308. The storage devices 306 can be a combination of volatile and non-volatile memory and can be at the same or different physical locations than the computing devices 302, 304. For example, the storage devices 306 can include any type of non-transitory computer readable medium capable of storing information, such as a hard-drive, solid state drive, tape drive, optical storage, memory card, ROM, RAM, DVD, CD-ROM, write-capable, and read-only memories.

[0045] The server computing device 302 can include one or more processors 310 and memory 312. The memory 312 can store information accessible by the processors 310, including instructions 314 that can be executed by the processors 310. The memory 312 can also include data 316 that can be retrieved, manipulated, or stored by the processors 310. The memory 312 can be a type of transitory or non-transitory computer readable medium capable of storing information accessible by the processors 310, such as volatile and non-volatile memory. The processors 310 can include one or more central processing units (CPUs), graphic processing units (GPUs), field- programmable gate arrays (FPGAs), and / or application- specific integrated circuits (ASICs), such as tensor processing units (TPUs).

[0046] The instructions 314 can include one or more instructions that, when executed by the processors 310, cause the one or more processors 310 to perform actions defined by the instructions 314. The instructions 314 can be stored in object code format for direct processing by the processors 310, or in other formats including interpretable scripts or collections of independent source code modules that are interpreted on demand or compiled in advance. The instructions 314 can include instructions for implementing a group persona generation system 318, which can correspond to the group persona generation system 200 as depicted in FIG. 2. The group persona generation system 318 can be executed using the processors 310, and / or using other processors remotely located from the server computing device 302.

[0047] The data 316 can be retrieved, stored, or modified by the processors 310 in accordance with the instructions 314. The data 316 can be stored in computer registers, in a relational or nonrelational database as a table having a plurality of different fields and records, or as JSON, YAML,proto, or XML documents. The data 316 can also be formatted in a computer-readable format such as, but not limited to, binary values, ASCII, or Unicode. Moreover, the data 316 can include information sufficient to identify relevant information, such as numbers, descriptive text, proprietary codes, pointers, references to data stored in other memories, including other network locations, or information that is used by a function to calculate relevant data.

[0048] The client computing device 304 can also be configured similarly to the server computing device 302, with one or more processors 320, memory 322, instructions 324, and data 326. The client computing device 304 can also include a user input 328 and a user output 330. The user input 328 can include any appropriate mechanism or technique for receiving input from a user, such as keyboard, mouse, mechanical actuators, soft actuators, touchscreens, microphones, and sensors.

[0049] The server computing device 302 can be configured to transmit data to the client computing device 304, and the client computing device 304 can be configured to display at least a portion of the received data on a display implemented as part of the user output 330. The user output 330 can also be used for displaying an interface between the client computing device 304 and the server computing device 302. The user output 330 can alternatively or additionally include one or more speakers, transducers or other audio outputs, a haptic interface or other tactile feedback that provides non-visual and non-audible information to the platform user of the client computing device 304.

[0050] Although FIG. 3 illustrates the processors 310, 320 and the memories 312, 322 as being within the respective computing devices 302, 304, components described herein can include multiple processors and memories that can operate in different physical locations and not within the same computing device. For example, some of the instructions 314, 324 and the data 316, 326 can be stored on a removable SD card and others within a read-only computer chip. Some or all of the instructions 314, 324 and data 316, 326 can be stored in a location physically remote from, yet still accessible by, the processors 310, 320. Similarly, the processors 310, 320 can include a collection of processors that can perform concurrent and / or sequential operation. The computing devices 302, 304 can each include one or more internal clocks providing timing information, which can be used for time measurement for operations and programs run by the computing devices 302, 304.

[0051] The server computing device 302 can be connected over the network 308 to a data center 332 housing any number of hardware accelerators 334. The data center 332 can be one of multiple data centers or other facilities in which various types of computing devices, such as hardware accelerators, are located. Computing resources housed in the data center 332 can be specified for deploying models, such as for group persona generation, as described herein.

[0052] The server computing device 302 can be configured to receive requests to process data from the client computing device 304 on computing resources in the data center 332. For example, the environment 300 can be part of a computing platform configured to provide a variety of services to users, through various user interfaces and / or application programming interfaces (APIs) exposing the platform services. As an example, the variety of services can include generating group personas. The client computing device 304 can transmit input data as part of a query for a task to generate a group persona. The group persona generation system 318 can receive the input data, and in response, generate output data including a response to the query including the generated group persona.

[0053] The server computing device 302 can maintain a variety of models in accordance with different constraints available at the data center 332. For example, the server computing device 302 can maintain different families for deploying models on various types of TPUs and / or GPUs housed in the data center 332 or otherwise available for processing.

[0054] FIG. 4 depicts a block diagram 400 illustrating one or more machine learning model architectures 402, more specifically 402A-N for each architecture, for deployment in a datacenter 404 housing a hardware accelerator 406 on which the deployed machine learning models 402 will execute, such as for the variety of services as described herein. The hardware accelerator 406 can be any type of processor, such as a CPU, GPU, FPGA, or ASIC such as a TPU.

[0055] An architecture 402 of a machine learning model can refer to characteristics defining the model, such as characteristics of layers for the model, how the layers process input, or how the layers interact with one another. The architecture 402 of the machine learning model can also define types of operations performed within each layer. One or more machine learning model architectures 402 can be generated that can output results, such as for group persona generation. Example model architectures 402 can correspond to generative models, such as language models, foundation models, and / or graphical models.

[0056] The machine learning models can be trained according to a variety of different learning techniques. Learning techniques for training the machine learning models can include supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning techniques. For example, training data can include multiple training examples that can be received as input by a model. The training examples can be labeled with a desired output for the model when processing the labeled training examples. The label and the model output can be evaluated through a loss function to determine an error, which can be back propagated through the model to update weights for the model. For example, a supervised learning technique can be applied to calculate an error between outputs, with a ground- truth label of a training example processed by the model. Any of a variety of loss or error functions appropriate for the type of the task the model is being trained for can be utilized, such as cross-entropy loss for classification tasks, or mean square error for regression tasks. The gradient of the error with respect to the different weights of the candidate model on candidate hardware can be calculated, for example using a backpropagation algorithm, and the weights for the model can be updated. The model can be trained until stopping criteria are met, such as a number of iterations for training, a maximum period of time, a convergence, or when a minimum accuracy threshold is met.

[0057] Referring back to FIG. 3, the devices 302, 304 and the data center 332 can be capable of direct and indirect communication over the network 308. For example, using a network socket, the client computing device 304 can connect to a service operating in the data center 332 through an Internet protocol. The devices 302, 304 can set up listening sockets that may accept an initiating connection for sending and receiving information. The network 308 can include various configurations and protocols including the Internet, World Wide Web, intranets, virtual private networks, wide area networks, local networks, and private networks using communication protocols proprietary to one or more companies. The network 308 can support a variety of short- and long-range connections. The short- and long-range connections may be made over different bandwidths, such as 2.402 GHz to 2.480 GHz, commonly associated with the Bluetooth® standard, 2.4 GHz and 5 GHz, commonly associated with the Wi-Fi® communication protocol; or with a variety of communication standards, such as the LTE® standard for wireless broadband communication. The network 308, in addition or alternatively, can also support wired connections between the devices 302, 304 and the data center 332, including over various types of Ethernet connection.

[0058] Although a single server computing device 302, client computing device 304, and data center 332 arc shown in FIG. 3, it is understood that the aspects of the disclosure can be implemented according to a variety of different configurations and quantities of computing devices, including in paradigms for sequential or parallel processing, or over a distributed network of multiple devices. In some implementations, aspects of the disclosure can be performed on a single device connected to hardware accelerators configured for processing machine learning models, or any combination thereof.

[0059] FIG. 5 depicts a flow diagram of an example process 500 for generating a group persona. The example process 500 can be performed on a system of one or more processors in one or more locations, such as the group persona generation system 200 as depicted in FIG. 2.

[0060] As shown in block 510, the group persona generation system 200 receives data associated with a description of characteristics for the group persona. The data can include generation data and / or refinement data, including detail for persona styles, images of one or more personas, and / or dependencies between one or more personas. The data can further include more detail about one persona over another persona.

[0061] As shown in block 520, the group persona generation system 200 generates a plurality of individual personas using a machine learning model based on the data associated with the description of characteristics for the group persona. The machine learning model can include a large language model, large foundation model, and / or a large graphical model. The machine learning model can be a general usage model or a fine-tuned model for group persona generation. The fine-tuned model can be trained on real-world or synthetic data associated with characteristics of various individual and group personas.

[0062] As shown in block 530, the group persona generation system 200 refines the plurality of individual personas using the machine learning model. The refinement can be iterative. The group persona generation system 200 can iteratively adjust the plurality of individual personas by adding further details to the personas or correcting details of the personas until reaching a stopping criterion.

[0063] As shown in block 540, the group persona generation system 200 generates inter-persona dynamics for the plurality of individual personas using the machine learning model based on the data associated with the description of characteristics for the group persona. The inter-persona dynamics and plurality of individual personas can form the group persona. The inter-personadynamics can include relationships among the plurality of individual personas. The group persona generation system 200 can generate the inter-persona dynamics by using a graph neural network to model the relationships among the plurality of individual personas. Nodes of the graph neural network can represent the individual personas and edges in the graph neural network can represent the relationships between the individual personas. The inter-persona dynamics can further include information spread and usage among the plurality of individual personas. The group persona generation system 200 can generate the inter-persona dynamics by using random graph processing to model the information spread and usage.

[0064] As shown in block 550, the group persona generation system 200 refines the inter-persona dynamics using the machine learning model. The refinement can be iterative. The group persona generation system 200 can iteratively adjust the inter-persona dynamics by adding further dynamics or correcting generated dynamics until reaching a stopping criterion.

[0065] As shown in block 560, the group persona generation system 200 outputs the group persona. The group persona can be implemented in various applications, such as simulating group characteristics in environments such as sales, advertising, employment, politics, teaching, gaming, etc.

[0066] Aspects of this disclosure can be implemented in digital electronic circuitry, in tangibly embodied computer software or firmware, and / or in computer hardware, such as the structure disclosed herein, their structural equivalents, or combinations thereof. Aspects of this disclosure can further be implemented as one or more computer programs, such as one or more modules of computer program instructions encoded on a tangible non-transitory computer storage medium for execution by, or to control the operation of, one or more data processing apparatus. The computer storage medium can be a machine-readable storage device, a machine -readable storage substrate, a random or serial access memory device, or combinations thereof. The computer program instructions can be encoded on an artificially generated propagated signal, such as a machinegenerated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus.

[0067] The term “configured” is used herein in connection with systems and computer program components. For a system of one or more computers to be configured to perform particular operations or actions means that the system has installed thereon software, firmware, hardware, or a combination thereof that cause the system to perform the operations or actions. For one or morecomputer programs to be configured to perform particular operations or actions means that the one or more programs include instructions that, when executed by one or more data processing apparatus, cause the apparatus to perform the operations or actions.

[0068] The term “data processing apparatus” or “data processing system” refers to data processing hardware and encompasses various apparatus, devices, and machines for processing data, including programmable processors, computers, or combinations thereof. The data processing apparatus can include special purpose logic circuitry, such as a field programmable gate array (FPGA) or an application specific integrated circuit (ASIC). The data processing apparatus can include code that creates an execution environment for computer programs, such as code that constitutes processor firmware, a protocol stack, a database management system, an operating system, or combinations thereof.

[0069] The term “computer program” refers to a program, software, a software application, an app, a module, a software module, a script, or code. The computer program can be written in any form of programming language, including compiled, interpreted, declarative, or procedural languages, or combinations thereof. The computer program can be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. The computer program can correspond to a file in a file system and can be stored in a portion of a file that holds other programs or data, such as one or more scripts stored in a markup language document, in a single file dedicated to the program in question, or in multiple coordinated files, such as files that store one or more modules, sub programs, or portions of code. The computer program can be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a data communication network.

[0070] The term “database” refers to any collection of data. The data can be unstructured or structured in any manner. The data can be stored on one or more storage devices in one or more locations. For example, an index database can include multiple collections of data, each of which may be organized and accessed differently.

[0071] The term “engine” refers to a software-based system, subsystem, or process that is programmed to perform one or more specific functions. The engine can be implemented as one or more software modules or components or can be installed on one or more computers in one ormore locations. A particular engine can have one or more computers dedicated thereto, or multiple engines can be installed and running on the same computer or computers.

[0072] The processes and logic flows described herein can be performed by one or more computers executing one or more computer programs to perform functions by operating on input data and generating output data. The processes and logic flows can also be performed by special purpose logic circuitry, or by a combination of special purpose logic circuitry and one or more computers.

[0073] A computer or special purpose logic circuitry executing the one or more computer programs can include a central processing unit, including general or special purpose microprocessors, for performing or executing instructions and one or more memory devices for storing the instructions and data. The central processing unit can receive instructions and data from the one or more memory devices, such as read only memory, random access memory, or combinations thereof, and can perform or execute the instructions. The computer or special purpose logic circuitry can also include, or be operatively coupled to, one or more storage devices for storing data, such as magnetic, magneto optical disks, or optical disks, for receiving data from or transferring data to. The computer or special purpose logic circuitry can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a Global Positioning System (GPS), or a portable storage device, e.g., a universal serial bus (USB) flash drive, as examples.

[0074] Computer readable media suitable for storing the one or more computer programs can include any form of volatile or non-volatile memory, media, or memory devices. Examples include semiconductor memory devices, e.g., EPROM, EEPROM, or flash memory devices, magnetic disks, e.g., internal hard disks or removable disks, magneto optical disks, CD-ROM disks, DVD- ROM disks, or combinations thereof.

[0075] Aspects of the disclosure can be implemented in a computing system that includes a back end component, e.g., as a data server, a middleware component, e.g., an application server, or a front end component, e.g., a client computer having a graphical user interface, a web browser, or an app, or any combination thereof. 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) and a wide area network (WAN), e.g., the Internet.

[0076] The computing system can include clients and servers. A client and server can be remote from each other and interact through a communication network. The relationship of client and server arises by virtue of the computer programs running on the respective computers and having a client-server relationship to each other. For example, a server can transmit data, e.g., an HTML page, to a client device, e.g., for purposes of displaying data to and receiving user input from a user interacting with the client device. Data generated at the client device, e.g., a result of the user interaction, can be received at the server from the client device.

[0077] Unless otherwise stated, the foregoing alternative examples are not mutually exclusive, but may be implemented in various combinations to achieve unique advantages. As these and other variations and combinations of the features discussed above can be utilized without departing from the subject matter defined by the claims, the foregoing description of the embodiments should be taken by way of illustration rather than by way of limitation of the subject matter defined by the claims. In addition, the provision of the examples described herein, as well as clauses phrased as "such as," "including" and the like, should not be interpreted as limiting the subject matter of the claims to the specific examples; rather, the examples are intended to illustrate only one of many possible embodiments. Further, the same reference numbers in different drawings can identify the same or similar elements.

Claims

CLAIMS1. A method for generating a group persona comprising: receiving, by one or more processors, data associated with a description of characteristics for the group persona; generating, by the one or more processors, a plurality of individual personas using a machine learning model based on the data associated with the description of characteristics for the group persona; generating, by the one or more processors, inter-persona dynamics for the plurality of individual personas using the machine learning model based on the data associated with the description of characteristics for the group persona, the inter-persona dynamics and plurality of individual personas forming the group persona; and outputting, by the one or more processors, the group persona.

2. The method of claim 1, wherein the machine learning model comprises at least one of a large foundation model, large language model, or large graphical model.

3. The method of claim 1, wherein the machine learning model is fine-tuned on at least one of real- world or synthetic data associated with characteristics of various individual and group personas.

4. The method of claim 1, wherein the data associated with the description of characteristics for the group persona comprises at least one of detail for one or more persona styles, images for one or more personas, or dependencies between one or more of the personas.

5. The method of claim 1, wherein the data associated with the description of characteristics for the group persona comprises more detail about an individual persona than another individual persona.

6. The method of claim 1, further comprising iteratively refining, by the one or more processors, one or more of the plurality of individual personas through outputting prompts.

7. The method of claim 1 , further comprising iteratively refining, by the one or more processors, the intcr-pcrsona dynamics through at least one of outputting prompts or simulating interactions between the plurality of individual personas.

8. The method of claim 1, wherein the inter-persona dynamics comprise relationships among the plurality of individual personas.

9. The method of claim 8, wherein generating the inter-persona dynamics for the plurality of individual personas comprises using a graph neural network to model the relationships among the plurality of individual personas, wherein nodes of the graph neural network represent individual personas and edges in the graph neural network represent relationships between the individual personas.

10. The method of claim 1, wherein the inter-persona dynamics comprise information spread and usage among the plurality of individual personas.

11. The method of claim 10, wherein generating the inter-persona dynamics for the plurality of individual personas comprises using random graph processing to model the information spread and usage.

12. A system comprising: one or more processors; and one or more storage devices coupled to the one or more processors and storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations for generating a group persona, the operations comprising: receiving data associated with a description of characteristics for the group persona; generating a plurality of individual personas using a machine learning model based on the data associated with the description of characteristics for the group persona; generating inter-persona dynamics for the plurality of individual personas using the machine learning model based on the data associated with the description of characteristicsfor the group persona, the inter-persona dynamics and plurality of individual personas forming the group persona; and outputting the group persona.

13. The system of claim 12, wherein the data associated with the description of characteristics for the group persona comprises at least one of detail for one or more persona styles, images for one or more personas, or dependencies between one or more of the personas.

14. The system of claim 12, wherein the operations further comprise iteratively refining one or more of the plurality of individual personas through outputting prompts.

15. The system of claim 12, wherein the operations further comprise iteratively refining the interpersona dynamics through at least one of outputting prompts or simulating interactions between the plurality of individual personas.

16. The system of claim 12, wherein the inter-persona dynamics comprise relationships among the plurality of individual personas.

17. The system of claim 16, wherein generating the inter-persona dynamics for the plurality of individual personas comprises using a graph neural network to model the relationships among the plurality of individual personas, wherein nodes of the graph neural network represent individual personas and edges in the graph neural network represent relationships between the individual personas.

18. The system of claim 12, wherein the inter-persona dynamics comprise information spread and usage among the plurality of individual personas.

19. The system of claim 18, wherein generating the inter-persona dynamics for the plurality of individual personas comprises using random graph processing to model the information spread and usage.

20. A non-transitory computer readable medium for storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations for generating a group persona, the operations comprising: receiving data associated with a description of characteristics for the group persona; generating a plurality of individual personas using a machine learning model based on the data associated with the description of characteristics for the group persona; generating inter-persona dynamics for the plurality of individual personas using the machine learning model based on the data associated with the description of characteristics for the group persona, the inter-persona dynamics and plurality of individual personas forming the group persona; and outputting the group persona.

Images