Connected Large Language Models With Segregated Private Data Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large language models require significant computational resources and may generate false information due to accessing sensitive data and the internet, leading to inefficiencies and security concerns.
Innovation Solution
Implement a system with a local large language model accessing a first limited private dataset and multiple external large language models, each with different private datasets, to segregate data access and prevent internet access, allowing for accurate and secure answer generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single large language model has access to all datasets including sensitive data, then the model can generate comprehensive answer data, but the computational resources required increase significantly and security risks increase
Solution Approach 1:
The patent divides the single large language model into multiple separate models, each with access to specific datasets. The local large language model has access to local datasets while external large language models have access to external datasets. This segmentation reduces the computational burden on any single model and improves security by limiting data access to only what is necessary for each model's function.
2Adaptability or versatility
If a large language model has access to all datasets, then it can provide comprehensive answers, but false information generation increases due to processing sensitive data
Solution Approach 1:
By segmenting the system into separate models with restricted data access, each model only processes data relevant to its designated datasets. This reduces the risk of false information generation from sensitive data while maintaining comprehensive answer capability through the coordinated operation of multiple specialized models.
Solution Approach 2:
Each large language model is assigned specific local datasets that are relevant to its function, rather than having access to all datasets. The local large language model accesses local datasets while external models access external datasets, ensuring that each model processes only appropriately labeled and relevant data, thereby reducing false information generation.
3Device complexity
If data is centralized in a single large language model, then data processing is simplified, but security and privacy protection are compromised
Solution Approach 1:
The patent implements segmentation by creating separate large language models for local and external data processing. The local large language model handles local datasets while external large language models handle external datasets, preventing any single model from accessing all data. This segmentation improves security and privacy protection while maintaining manageable system complexity through clear data access boundaries.
Data Source
AI summary
Methods, systems, apparatuses, and non-transitory computer-readable media are provided for providing answer data through multiple connected large language models. Operations may include receiving, through a graphical user interface associated with a local large language model having access to a first limited private dataset but not a second limited private dataset, an input from a user device, identifying, based on the input, an external large language model from among a plurality of external large language models, transmitting the input to the external large language model, receiving, from the external large language model, the answer data responsive to the input, generating, by the local large language model, response data based on the answer data, and outputting the response data at the user device.