Related Experiment Video
Updated: Jun 29, 2025

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
Powerful tool or too powerful? Early public discourse about ChatGPT across 4 million tweets
Reuben Ng1,2, Ting Yu Joanne Chow1
1Lee Kuan Yew School of Public Policy, National University of Singapore, Singapore, Singapore.
Background:
This paper investigates initial exuberance and emotions surrounding ChatGPT's first three months of launch (1 December 2022-1 March 2023). The impetus for studying active discussions surrounding its implications, fears, and opinions is motivated by its nascent popularity and potential to disrupt existing professions; compounded by its significance as a crucial inflexion point in history. Capturing the public zeitgeist on new innovations-much like the advent of the printing press, radio, newspapers, or the internet-provides a retrospective overview of public sentiments, common themes, and issues.
Objectives:
Since launch, few big data studies delved into initial public discourse surrounding the chatbot. This report firstly identifies highest-engagement issues and themes that generated the most interaction; secondly, identifies the highest-engaged keywords on both sides of the sentiment valence scale (positive and negative) associated with ChatGPT.
Methods:
We interrogate a large twitter corpus (n = 4,251,662) of all publicly available English-language tweets containing the ChatGPT keyword. Our first research aim utilizes a prominent peaks model (upper-quartile significance threshold of prominence>20,000). Our second research aim utilized sentiment analysis to identify, week-on-week, highest-frequency negative, and positive keywords and emojis.
Results:
Six prominent peaks were identified with the following themes: 'hype and hesitance', 'utility and misuse in professional and academic settings', 'demographic bias', 'philosophical thought experiments on morality' and 'artificial intelligence as a mirror of human knowledge'. Of high-frequency valence, negativity included credibility concerns, implicit bias, environmental ethics, employment rights of data annotators and programmers, the ethicality of neural network datasets. Positivity included excitement over application, especially in coding, as a creative tool, education, and personal productivity.
Conclusions:
Overall, sentiments and themes were double-edged, expressing excitement over this powerful new tool and wariness toward its potential for misuse.
More Related Videos
08:53Integrating Computerized Linguistic and Social Network Analyses to Capture Addiction Recovery Capital in an Online Community
Published on: May 31, 2019
09:40Measuring Neural and Behavioral Activity During Ongoing Computerized Social Interactions: An Examination of Event-Related Brain Potentials
Published on: November 15, 2014
Related Concept Videos
Group Polarization
Social Exchange Theory
Genetic Lingo
Genome-wide Association Studies-GWAS
GWAS does not require the identification of the target gene involved in...
Real Time RT-PCR
The real-time quantification of the number of amplified products is...
GTPases and their Regulation
Large G-proteins,...