Virtual Assistant Machine Learning with Mesh Language Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI assistant models struggle to provide a human parity experience due to insufficient training data, especially in specialized contexts, and lack a continuous learning mechanism to adapt to changing inputs.
Innovation Solution
A virtual assistant platform utilizing a smart mesh network that combines a generic language model with client-specific and mesh-specific models, enabling continuous learning and data sharing across multiple assistants to enhance training efficiency and adaptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a centralised learning model is used to train a common code base on interactions from a large set of users, then the model can effectively handle generic queries common to all users, but it cannot accomplish specialist tasks appropriate for certain subsets of users
Solution Approach 1:
The patent segments the learning model into multiple specialized models, each trained on interactions from specific user subsets or organizations. This allows each model to become an expert in its domain while the system as a whole can handle diverse specialist tasks through model selection or ensemble methods.
Solution Approach 2:
The patent creates a universal framework that can deploy different specialized models for different tasks. The system maintains a registry of multiple models and can dynamically select or combine them based on the query type, achieving both generic and specialist task handling through a single multi-functional platform.
2Adaptability or versatility
If a Stand Alone model is trained locally on an organisation's limited pool of data to provide specialist tasks, then the model can be customized for specific organizational needs, but it typically fails to offer a human parity experience due to insufficient training data
Solution Approach 1:
The patent merges data from multiple organizations or user groups to create enriched training datasets. By combining limited local data with external data from similar contexts, the model achieves sufficient training data volume and diversity to provide human parity experience while maintaining organizational specificity.
Solution Approach 2:
The patent introduces an intermediary data sharing layer that allows organizations to contribute anonymized interaction data to a collective pool. This intermediary mechanism enables each organization to benefit from aggregated data from others, improving model performance without exposing sensitive organizational information.
3Ease of manufacture
If a Stand Alone model is trained once and then deployed without further training after going to production, then the deployment process is simple, but the model is only ever as effective as it was at deployment time and cannot adapt to constantly new and changing inputs
Solution Approach 1:
The patent implements continuous learning by enabling the model to ongoingly ingest new interaction data from production environments. The system maintains data collection and model retraining capabilities that operate continuously, ensuring the model adapts to new patterns and maintains effectiveness over time without requiring full redeployment cycles.
Solution Approach 2:
The patent incorporates feedback mechanisms where model performance is continuously monitored and evaluation results are used to trigger retraining cycles. This feedback loop ensures the model adapts to changing inputs by identifying performance gaps and initiating targeted retraining with relevant new data.
Data Source
AI summary
A virtual assistant platform implemented by a computer system comprising: one or more hardware processors configured to execute computer readable instructions; one or more memory storing the instructions; a mapping data structure stored in the one or more memory, the mapping data structure mapping a plurality of intents to respective client specific actions; a network interface configured to receive a query from a user device operating in a client specific communication session with a virtual assistant in a first context, the instructions when executed providing: an AI language model comprising a client specific language model, the client specific language model having been trained on client specific data, and a mesh language model, the mesh language model having been trained on mesh specific data, the mesh specific data having been received by operating multiple virtual assistants in the first context, the AI language model being responsive to the query to generate an intent; a mapping function to apply the intent to the mapping data structure and access a corresponding client specific action for delivery of a response to the user device; and a transmission function to transmit the response to the user device.