Evaluasi Efisiensi Sumber Daya pada Proses Prompting AI Berdasarkan Konsep Green Computing

Bagus Aufa Ariq Armeidi; M. Krisna Pramudya

doi:10.52303/jb.v8i1.221

Bagus Aufa Ariq Armeidi Universitas Multi Data Palembang
M. Krisna Pramudya Universitas Multi Data Palembang

DOI: https://doi.org/10.52303/jb.v8i1.221

Keywords: Green Computing, Prompting AI, Energy Efficiency, Resource Management, Artificial Intelligence

Abstract

The rapid advancement of artificial intelligence (AI) technology has triggered a significant surge in computational resource consumption, particularly in the inference phase driven by text prompting. This study aims to evaluate the resource efficiency of various AI prompting techniques based on Green Computing principles, which emphasize minimizing energy consumption and maximizing the computational sustainability of digital systems. The research employs a qualitative descriptive literature review method, collecting data from 15 scientific publications sourced from Google Scholar, IEEE Xplore, and Semantic Scholar databases, covering the period 2020–2025. The analysis demonstrates that prompting technique selection directly influences token count, response latency, and energy consumption. Zero-shot prompting exhibits the lowest energy footprint (0.38 mWh per query) while Chain-of-Thought and ReAct prompting, although yielding higher accuracy (91–93%), consume up to five times more energy. The findings indicate that applying prompt compression, context pruning, and model-right-sizing strategies, aligned with Green Computing principles can reduce energy usage by 40–60% without substantially sacrificing output quality. This research contributes a practical framework for designing energy-aware AI prompting strategies suitable for sustainable computing environments.

References

[1] T. Brown et al., “Language Models are Few-Shot Learners,” Adv. Neural Inf. Process. Syst., vol. 33, 2020.
[2] E. Strubell, A. Ganesh, and A. McCallum, “Energy and Policy Considerations for Modern Deep Learning Research,” Proc. AAAI Conf. Artif. Intell., vol. 34, 2020.
[3] D. Patterson et al., “Carbon Emissions and Large Neural Network Training,” arXiv Prepr. arXiv2104.10350, 2021.
[4] P. Liu, W. Yuan, J. Fu, Z. Jiang, H. Hayashi, and G. Neubig, “Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing,” ACM Comput. Surv., vol. 55, 2023.
[5] J. Wei et al., “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models,” Adv. Neural Inf. Process. Syst., vol. 35, 2022.
[6] T. Hao, J. Xu, and Q. Li, “Green Computing: A Survey of Current Trends and Future Directions,” IEEE Access, vol. 10, 2022.
[7] A. Lacoste, A. Luccioni, V. Schmidt, and T. Dandres, “Quantifying the Carbon Emissions of Machine Learning,” arXiv Prepr. arXiv1910.09700, 2022.
[8] H. Jiang, Q. Wu, C.-Y. Lin, Y. Yang, and L. Qiu, “LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models,” Proc. EMNLP 2023, 2023.
[9] A.-L. Ligozat, J. Lefevre, A. Bugeau, and J. Combaz, “Unraveling the Hidden Environmental Impacts of AI Solutions for Environment,” Sustainability, vol. 14, 2022.
[10] Z. Gou et al., “CRITIC: Large Language Models can Self-Correct with Tool-Interactive Critiquing,” arXiv Prepr. arXiv2305.11738, 2023.
[11] R. Schwartz, J. Dodge, N. A. Smith, and O. Etzioni, “Green AI,” Commun. ACM, vol. 63, 2020.
[12] N. Bannour, S. Ghannay, A. Névéol, and A.-L. Ligozat, “Evaluating the Carbon Footprint of NLP Methods: A Survey and Analysis of Existing Tools,” Proc. 2nd Work. Simple Effic. Nat. Lang. Process., 2021.
[13] S. Yao et al., “ReAct: Synergizing Reasoning and Acting in Language Models,” Int. Conf. Learn. Represent., 2023.
[14] K. Lottick, S. Susai, S. A. Friedler, and J. P. Wilson, “Energy Usage Reports: Environmental Awareness as Part of Algorithmic Accountability,” NeurIPS Work. Tackling Clim. Chang. with AI, 2020.
[15] S. Lim and A. Kim, “Towards Energy-Efficient Inference in Large Language Models: Techniques and Benchmark,” IEEE Trans. Sustain. Comput., vol. 9, 2024.