Workflow of Multimodal GPT in Circular Economy Modeling and Simulation.
Explanation: This step-by-step workflow chart illustrates how multimodal GPT may enhance Material Flow Analysis (MFA) and System Dynamics (SD) by recognizing and interpreting system model diagrams, facilitating accurate model construction and enabling the set-up of dynamic simulations of Taiwan’s circular economy scenarios.
As the consumption of global resources mounts and human pressures on the environmental intensify, 1 the problem of balancing economic growth with sustainability 2 has become a shared challenge for governments, industries, and academia. The “circular economy” model is widely touted as a promising solution, yet identifying optimal strategies amid complex policies, industrial data, and resource flows in diverse contexts remains a formidable task.
A cross-disciplinary research team at National Taiwan University (NTU) has made a significant breakthrough by applying multimodal Generative Pre-trained Transformer (GPT) models to the analysis of circular economy transformation. Their study, Application and Scenario Simulation of Multimodal GPT in Circular Economy Transformation: A Case Study of Taiwan’s Material Flow Data, was recently published in the prestigious journal Resources, Conservation & Recycling.
Working with Taiwan’s available material flow data and policies from 2013 to 2022, the team integrated Material Flow Analysis (MFA) and System Dynamics (SD) models with GPT’s multimodal capabilities—interpreting text, data, and images—to set up a policy scenario simulation framework. In this way, they devised an end-to-end process, involving scenario design, data computation, and visualization of results that will enable official policymakers, industry stakeholders, as well as the general public to understand the long-term impacts of policy changes.
In Scenario 9 (Population Stabilization and Resource Efficiency), the simulations project that Taiwan’s circular material use rate will increase from 22% in 2022 to 29% by 2030, while resource productivity will rise from NT$65 to NT$88 per kilogram. These predicted gains indicate both more efficient use of materials and also greater economic value created per unit of resource—confirming the dual environmental and economic benefits of well-designed policy interventions.
This research further underscores how AI can serve as a powerful strategic tool for sustainability—reducing barriers to cross-disciplinary collaboration and making policy planning more scientific and transparent. The team plans to extend the application of multimodal GPT to such areas as climate change, waste management, and circular industry policy development. They welcome collaborations with academic, industrial, and policy partners worldwide to co-create innovative approaches toward a realizing more sustainable and environmentally-friendly future.
1 Human activities place pressure on the environment through habitat destruction (deforestation, urbanization), pollution (air, water, soil, and plastic), and the over-exploitation of resources (overharvesting, overfishing).2 Both resource sustainability and environmental sustainability.
Sample Scenario Simulation Output.
Explanation: Line graphs of a GPT-assisted scenario simulation (Scenario 9: Population Stabilization and Resource Efficiency), showing projected increases in Taiwan’s circular material use rate and resource productivity.
Click or Scan the QR code
to read the full article
published in Resources,
Conservation & Recycling.
