Generative AI Occupant Assistant for Building Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building management systems face challenges in precisely generating relevant data for specific conditions due to limitations in predictive nature of language models, which can produce incorrect, imprecise, or irrelevant text data, requiring manual user intervention and being resource-intensive for processing equipment operation data.
Innovation Solution
Implementing machine learning models, including generative AI systems, to capture unstructured data from various sources, process it, and generate accurate outputs in structured formats for applications like equipment servicing, using automated and expert-based thresholds to improve data quality and update training models in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If language models are used to generate text data for building management, then conversational interfaces and messaging capabilities are enabled, but the predictive nature of language models produces incorrect, imprecise, or irrelevant data requiring manual intervention
Solution Approach 1:
The patent introduces an intermediary processing layer between the language model and the building management system. This layer includes data validation mechanisms, confidence threshold checks, and automated correction protocols that filter and verify the generated text data before it is used for operational decisions, thereby maintaining reliability while preserving conversational interface capabilities
Solution Approach 2:
The system implements feedback loops where generated text data is validated against actual building management data, and discrepancies are used to retrain and improve the language model. Manual corrections and validations are fed back into the system to continuously enhance accuracy, reducing the need for ongoing manual intervention
2Adaptability or versatility
If language models process equipment operation data, then conversational responses can be generated, but the processing is resource-intensive
Solution Approach 1:
The patent segments the data processing workload by dividing equipment operation data into distinct categories and processing them through specialized modules before being passed to the language model. This segmentation allows only relevant, pre-processed data to be used for conversational responses, significantly reducing computational resource consumption while maintaining versatility
Solution Approach 2:
The system performs preliminary processing of equipment operation data, including filtering, aggregation, and relevance assessment, before the data is input to the language model. This preliminary action reduces the volume and complexity of data requiring intensive computational processing, thereby reducing energy consumption while preserving the model's ability to generate appropriate responses
3Productivity
If generative AI models are trained on preexisting data sets, then the models can generate responses, but the models cannot produce new data not present in the training data
Solution Approach 1:
The patent implements dynamic training mechanisms where the generative AI model is continuously retrained on new building management data as it becomes available. This dynamic approach allows the model to adapt to new equipment, new operational patterns, and new data formats, enabling it to generate novel responses while maintaining high productivity through efficient processing of familiar data types
Data Source
AI summary
A method can include providing, by one or more processors, a digital occupant assistant application for occupants of a commercial building and dynamically generating, by the one or more processors using a generative artificial intelligence (AI) model, data to present to an occupant via the digital occupant assistant application. The generative AI model can dynamically generate the data based on at least one of a prompt from the occupant provided via an input interface of the digital occupant assistant application or context relating to at least one of the occupant, the building, a space of the building, an event relating to the building, one or more other occupants, or building equipment or other assets of the building.