Natural Language Agent Session History for Module Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated virtual agents require significant configuration and computational resources to adapt to updates in the environment, such as changes in application modules, leading to ongoing training overhead and inefficiencies.
Innovation Solution
The implementation of natural language models trained to interact with users and environments in a goal-directed manner, using an interpreter to generate commands and learn about available functionality, allowing the model to adapt to updates by discarding outdated information and re-learning current information at the start of each session.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If automated virtual agents are configured to interact with updated environment modules, then the agents can service user needs with new functionality, but significant computational resources and configuration effort are required for retraining
Solution Approach 1:
The system performs preliminary actions by having the virtual agent proactively query and learn about available modules and their functionalities at the beginning of each session. The agent executes information-gathering commands to discover module documentation, function signatures, and usage patterns before needing to use them, thereby preparing itself in advance for potential updates without requiring full retraining
Solution Approach 2:
The virtual agent serves itself by autonomously discovering and learning about environment modules through self-executed information queries. Instead of requiring external retraining configurations, the agent independently gathers information about available functionality by executing commands to access module documentation and metadata, then uses this self-acquired knowledge to adapt to updates
2Reliability
If the virtual agent maintains a comprehensive history of all interactions, then it can improve its responses through context, but computational requirements for running the model increase
Solution Approach 1:
The system extracts only the essential and relevant information from the interaction history that is necessary for maintaining context, rather than preserving the complete history. By selectively extracting key contextual elements and discarding redundant information, the system maintains reliable responses while reducing the computational burden of processing and storing extensive historical data
Data Source
AI summary
A trained natural language model is provided that uses an input session history to generate outputs to an interpreter. Outputs to the interpreter, and inputs responsively received therefrom, are added to the history to generate additional model outputs as the history is updated. The model is trained to engage in goal-oriented dialog with the interpreter and with the user (optionally through interpreter function calls) to identify the user's goals, to learn information about modules, functions, and methods available in the interpreter that are relevant to the user's goals, and to execute function calls and/or commands, based on the learned information, that accomplish the user's goals. The use of a history that may be completely blank at the beginning of the session reduces the computational requirements of running the model, as well as allowing the model to ‘update’ itself as the available modules are update, added, or removed.