AI usefulness in systems modelling and simulation : GPT-4 application

Abstract

In this study, we investigate the potential of artificial intelligence (AI), specifically Generative Pre-trained Transformer 4 (GPT-4), to accelerate the development of system dynamics simulations within the broader context of systems engineering. The research aims to uncover the opportunities and limitations of leveraging AI to assist humans in constructing and refining system dynamics models. Through a systematic iterative process, GPT-4 was engaged in tasks such as creating, expanding, and stabilising simulations, identifying errors, generating expansion ideas, and converting models to Python code. Our findings reveal that GPT-4, while not flawless, can significantly enhance the modelling process, reduce human error, and expedite learning. This paper critically examines the role of AI in model development, emphasising the continued importance of human expertise in the evaluation and testing of simulations. Ultimately, we argue for a symbiotic relationship between AI and human modellers, harnessing the power of GPT-4 to augment human capabilities and advance the fields of system dynamics and systems engineering.

In hierdie studie ondersoek ons die potensiaal van kunsmatige intelligensie (KI), spesifiek Generative Pre-trained Transformer 4 (GPT)-4), om die ontwikkeling van stelseldinamika-simulasies binne die breër konteks van stelselingenieurswese. Die navorsing het ten doel om die geleenthede en beperkings van die gebruik van KI te ontbloot om mense te help om stelseldinamika-modelle te konstrueer en te verfyn. Deur 'n sistematiese, iteratiewe proses was GPT-4 besig met take soos die skep, uitbreiding en stabilisering van simulasies, identifisering van foute, generering van uitbreidingsidees en omskakeling van modelle na Python-kode. Ons bevindinge toon dat GPT-4, hoewel dit nie foutloos is nie, die modelleringsproses aansienlik kan verbeter, menslike foute kan verminder en leer bespoedig. Hierdie artikel doen ‘n kritiese ondersoek na die rol van KI in modelontwikkeling, met die klem op die voortgesette belangrikheid van menslike kundigheid in die evaluering en toetsing van simulasies. Uiteindelik argumenteer ons vir 'n simbiotiese verhouding tussen KI en menslike modelleerders, wat die krag van GPT-4 benut om menslike vermoëns te verbeter en om die velde van stelseldinamika en stelselingenieurswese te bevorder.