# Comparing ChatGPT, Microsoft Copilot, and Google Gemini for specific article retrieval and summarization

> A comparative analysis of how different AI models handle article retrieval and summarization tasks based on the ÄLLITÄ project at Savonia University of Applied Sciences.

## Methodology and Comparison
In this article, we tested three AI models—Enterprise Microsoft Copilot (Savonia’s access), ChatGPT 5.0 (free edition), and Gemini 2.5 Pro—using the prompt: *"search articles from savonia.fi related to ällitä project only and write a story-based summary of 500 words."*

*   **Microsoft Copilot:** Produced a 471-word summary with a title but struggled to extract all relevant articles and information.
*   **ChatGPT:** Produced a 740-word summary. It successfully extracted all project articles and included embedded links, but failed to adhere to the 500-word limit.
*   **Gemini:** Produced a 592-word summary. It extracted most of the information and provided the most structured response, organizing content by Work Packages (WPs).

## The ÄLLITÄ Project Overview
The ÄLLITÄ project, officially titled “Smart heating systems in a climate-friendly way by utilizing artificial intelligence,” is an EU co-financed initiative led by Savonia University of Applied Sciences. The project focuses on two primary objectives: optimizing building energy systems and leveraging generative AI to automate routine tasks.

## Energy and IoT Innovation
The project utilized IoT sensors across Savonia’s Varkaus and Kuopio campuses to monitor electricity consumption, building occupancy, and weather conditions. 

*   **Predictive Modeling:** Researchers analyzed 16 months of electricity data alongside weather metrics (temperature, humidity, snow depth). By incorporating time-based cyclic features and lag features, they significantly improved the accuracy of linear regression models used to forecast energy demand.
*   **Solar Energy:** A 2 kW solar PV system at the Kuopio campus served as a laboratory. Data collected from April to September 2024 showed peak production in May, with daily outputs peaking between 11 AM and 3 PM. These insights are being used to develop AI-driven models to predict renewable energy output.
*   **Building Automation:** The integration of hardware and AI enables dynamic energy optimization, moving toward smarter, more sustainable campus operations.

## Generative AI and Digital Empowerment
Work Package 3 focused on using AI to streamline business operations and reduce repetitive tasks:

*   **Document Summarization:** The team developed a web application using open-source models like Facebook’s BART to summarize lengthy PDFs, assisting researchers in managing data.
*   **Software Development:** Generative AI models, including Meta’s Llama-3.3-70B-Instruct, were used to generate frontend and backend code, debug web layouts, and write code for microcontrollers.
*   **3D Generation:** The project integrated Microsoft’s TRELLIS, allowing users to generate 3D models from text prompts, facilitating rapid product design and concept ideation.

## Conclusion
The ÄLLITÄ project serves as a strategic effort to strengthen the digital competence of the Northern Savo region. By combining renewable energy forecasting, building automation, and AI-based software tools, Savonia is building the infrastructure for a future where energy efficiency and productivity are integrated.

## Authors
*   **Shahbaz Baig**, RDI Specialist, DigiCenter, Savonia-ammattikorkeakoulu, shahbaz.baig@savonia.fi
*   **Mika Leskinen**, RDI Specialist, DigiCenter, Savonia-ammattikorkeakoulu, mika.leskinen@savonia.fi
*   **Aki Happonen**, Digital Development Manager, DigiCenter, Savonia-ammattikorkeakoulu, aki.happonen@savonia.fi
*   **Laura Leppänen**, RDI Specialist, Savonia-ammattikorkeakoulu Oy, laura.leppanen@savonia.fi