Security Challenges and Reflections on Large Models-Reference-Cited by-同舟云学术

Security Challenges and Reflections on Large Models

Published:2024-08-27 Issue:2 Volume:9 Page:1-3
ISSN:2832-6024
Container-title:Frontiers in Computing and Intelligent Systems
language:
Short-container-title:FCIS

Author:

Huang Xuan,Huang Linyi,Tong Guowei,Zhou Xundao,Luo Jianheng

Abstract

The rapid development of large-scale AI models has revolutionized the technology industry, offering unprecedented opportunities for innovation across various sectors. This paper discusses the emergence of the "Hundred Model War" and the significant growth in large models, highlighting the potential for transformative applications in vertical fields such as automotive, medical, and finance. However, we also identify significant challenges, including safety ethics, governance systems, and the vulnerability of models to malicious attacks. The paper concludes with a call for the establishment of a comprehensive ethical governance system, improved safety supervision mechanisms, and the development of a public technology resource support platform to ensure the sustainable and healthy development of AI technologies.

Publisher

Darcy & Roy Press Co. Ltd.

Reference8 articles.

1. [1] OpenAI. (2024). GPT-4o: A Multimodal AI Model. [Online] Available: https://openai.com/2024/gpt-4o [Accessed: July 2024].

2. [2] Zhang Bolin; Tu Zhiying; Hang Shaoshi, Chu Dianhui, Xu Xiaofei, “Conco-ERNIE: Complex [2] User Intent Detect Model for Smart Healthcare Cognitive Bot”, ACM Transactions on Internet Technology, Vol.1, 2023.

3. [3] Pan Zhang, Xiaoyi Dong, Yuhang Zang, et al. “InternLM-XComposer-2.5: A Versatile Large Vision Language Model Supporting Long-Contextual Input and Output”, arXiv: 2407. 03320.

4. [4] Guohai Xu, Jiayi Liu, Ming Yan, et al. “CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility”, arXiv:2307.09705.

5. [5] H Dan, C Burns, S Basart, et al. “Measuring Massive Multitask Language Understanding”, Proceedings of the 9th International Conference on Learning Representations (ICLR 2021), 2021.