How we Differ - The Challenge of Building Digital Twins
Digital twins (DT) have enormous promise for automating and optimizing assets, processes, and complex operations. However, developing DTs remains a challenging and costly task in many cases.
The difficulty can be summed up in defining and updating the model used by a DT to capture the relationships between the thousands of parameters required to model a supply chain network, a production line, or a human patient.
The Social AI Platform Approach
The Social AI Platform takes a unique and new approach to modeling complex processes and interactions.
To begin, after defining the model scope in terms of the many parameters involved in a process or scenario, we apply an inference method to determine the significance of each parameter for predicting a specific event in the scenario. The result is what we call a Reference Events Model (REM).
The REM is then used by the various Information Digital Twins (IDT) to monitor their part of the scenario (or the entire scenario) and predict the onset of the events of interest. Up until this point, no machine learning has been used, only our proprietary Semantic Algorithm.
After ensuring the consistency and reliability of the various predictions, machine learning is applied to the REM and every IDT to learn patterns and correlations among the thousands of involved parameters and the corresponding events. The insights gained from applying machine learning increase the speed and accuracy of event predictions but mainly reduce the model's complexity.
Finally, the learned patterns enable the IDTs to automate decisions and ensure effective scenario performance.
The Social AI Platform Overview
The Social AI Platform is the realization of an information architecture framework for modeling, managing, and automating hundreds of highly complex scenarios and procedures.
The key to achieving this level of coordination is to represent all involved data according to a consistent data structure, which is defined by the platform information architecture.
In other words, the information architecture captures the communication - exchange of information - between the various scenarios and the users in a consistent manner.
Reference Events Model (REM)
The REM can be seen to serve one purpose: providing the number of bits required to control the event it represents. The greater the number, the more difficult it is to manage the event and vice versa. The probability of the various input and output parameters provide the insights necessary to control the event.
The Reference Events Model (REM) parametrizes a process or scenario. The parameters are classified as input, context, and output. In the above example, a REM is defined to manage specific risks associated with an ICU patient.
The input and context parameters provide information about the medications and patient-specific conditions and history. In contrast, the output parameters provide information about the various conditions observed for the patient.
Using inference from historical data, the REM then calculates how well a parameter predicts the risks monitored. Each patient is provided their own Patient-Human Digital Twin, or Patient-HDT, which relies on the various probabilities from the REM to predict the onset of any of the risks in the scope of the REM.
In a broad sense, the model inputs are the parameters that can be adjusted to govern the model-managed event.
Information Digital Twin (IDT)
The IDT's primary goal is to improve its user capacity to predict its environment and be predicted by it (regardless of the nature of the user activities), which significantly improves its user's ability to collaborate and automate their interactions with their environment.
To model and automate agent-environment interactions (i.e., communication), the Information Digital Twin relies on three major components. A representation component contains the model of the agent's interactions it supports. A learning component to learn and update the model according to the specifics and behavior of the agent. A monitoring and control component to evaluate the agent's behavior and interact with it either in the form of sending recommendations or taking over some of the actions.
The IDT representation is initialized by the REM that manages the scenario involving the various agents. The IDT representation is thus a modified form of the REM, it has a similar structure, but the parameter probabilities are updated according to the agent behavior. The IDT uses the representation to monitor and predict certain events, learn its agent interaction patterns, update the REM, provide recommendations to its agent or take over some of the agent's decisions.
It is critical to state that an IDT can support many REMs simultaneously. Suppose an agent is involved with various scenarios across multiple domains. In that case, its IDT can connect to the different Social AI Platforms simultaneously to track and predict events along with different scenarios. The IDT's model, in this case, is a composite of the various REMs it supports. This allows the IDT to learn cross-dependencies and correlations between scenarios rather than only within them. For example, suppose an IDT is used to track a person's health risks. In that case, it will be able - if desired by its user - to learn parameters from others across various aspects of a person's lifestyle that may influence their disease risk, e.g., the impact of their work or driving habits on their health.
The IDT observes and learns through its agent's interaction (communication) with the environment. As a result, the IDT is unconcerned with the nature of its agent, whether human or machine or any other entity of interest (e.g., a shipment), as long as their interactions can be captured and modeled using the data structures of the Social AI Platform. However, we refer to the IDT as a Human Digital Twin if it is employed to assist a human (HDT), as most of our use cases involve enabling humans in their interactions with their environment.
The Semantic Algorithm
The Semantic Algorithm is Bayesian-based. As such, it calculates the relevance of a parameter to various specific events of interest. As an event is usually dependent on a large number of parameters, the algorithm also enables an Information Digital Twin (IDT) to calculate the joint probabilities of multiple input and output parameters to predict the occurrence of an event. The algorithm can be configured to selectively update the Reference Events Model (REM) with the probabilities of specific parameters and/or events.