Exploring AI Ethics of ChatGPT: A Diagnostic Analysis

By Terry Yue Zhuo, Yujin Huang, Chunyang Chen, Zhenchang Xing. 2023

262 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Health Informatics, Safety Research, Safety, Robustness & Interpretability, AI Ethics, Fairness & Society

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Health Informatics

In Safety Research

In Safety, Robustness & Interpretability

In AI Ethics, Fairness & Society

Citations in the corpus

Llama 2: Open Foundation and Fine-Tuned Chat Models (2023, background): “Investigations into red teaming reveal specific challenges in tuned LLMs, with studies by Ganguli et al. (2022) and Zhuo et al. (2023) showcasing a variety of successful attack types and their effects on the generation of harmful content.”
Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey (2022, background): “works have outlined these risks (Maynez et al., 2020; Sheng et al., 2021; Weidinger et al., 2022; Zhuo et al., 2023), proposed taxonomies (Weidinger et al.”
Can Large Language Models Transform Computational Social Science? (2023, unknown): “C L ] 1 2 A pr 2 02 3 Zhuo et al., 2023; Goyal et al., 2022).”
ChatGPT and Open-AI Models: A Preliminary Review (2023, background): “General Current benchmarks on ChatGPT cannot adequately address a significant proportion of ethical concerns [54].”
A Survey on Evaluation of Large Language Models (2023, methodology): “[251] used conventional testing sets and metrics [35, 50, 144] to perform a systematic evaluation of ChatGPT’s toxicity and social bias, finding that it still exhibits noxious content to some extend.”
Driving with LLMs: Fusing Object-Level Vector Modality for Explainable Autonomous Driving (2023, background): “By introducing LLMs into autonomous driving we are inheriting their ethical implications [50].”
Recommender Systems in the Era of Large Language Models (LLMs) (2023, unknown): “LLMs have been proven to advance recommender systems in various aspects, but they are also highly vulnerable to adversarial perturbations ( i.e. , minor changes in the input) that can compromise the safety and robustness of their uses in safety-critical applications [102].”
Can ChatGPT Understand Too? A Comparative Study on ChatGPT and Fine-tuned BERT (2023, background): “There is growing interest in exploring the capabilities, applications, ethics, and failures of ChatGPT (Jiao et al., 2023; Bang et al., 2023; Qin et al., 2023; Zhuo et al., 2023; Wang et al., 2023).”
Summary of ChatGPT-Related Research and Perspective Towards the Future of Large Language Models (2023, background): “[112] comprehensively analyzed the moral hazard, bias, reliability, robustness, and toxicity of ChatGPT from four perspectives.”
Summary of ChatGPT/GPT-4 Research and Perspective Towards the Future of Large Language Models (2023, background): “[106] comprehensively analyzed the moral hazard, bias, reliability, robustness, and toxicity of ChatGPT from four perspectives.”
Toxicity in ChatGPT: Analyzing Persona-assigned Language Models (2023, result): “Contrary to the results of prior studies (Zhuo et al., 2023), we find that CHATGPT can be consistently toxic about a wide range of topics when it is assigned a persona.”
On the Robustness of ChatGPT: An Adversarial and Out-of-distribution Perspective (2023, background): “The evaluations on ethics are done in [62].”
Large Language Model as Attributed Training Data Generator: A Tale of Diversity and Bias (2023, background): “Notably, the prevailing approach employs a simple class-conditional prompt for querying LLMs during data generation, potentially limiting the diversity of the generated data [7, 51, 60] and inheriting systematic biases inherent in LLMs [65, 21].”
Unifying Large Language Models and Knowledge Graphs: A Roadmap (2023, background): “The hallucination problem in LLMs [246], which generates factually incorrect content, significantly hinders the reliability of LLMs.”
ChatGPT: Jack of all trades, master of none (2023, background): “Such research has already been conducted to some extent and multiple interesting findings about ethical aspects of ChatGPT’s performance in comparison to different large language models can be found in [52].”