Cogent Business & Management

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/oabm20

ChatGPT: A brief narrative review

Bulbul Gupta, Tabish Mufti, Shahab Saquib Sohail & Dag Øivind Madsen

To cite this article: Bulbul Gupta, Tabish Mufti, Shahab Saquib Sohail & Dag Øivind Madsen
(2023) ChatGPT: A brief narrative review, Cogent Business & Management, 10:3, 2275851, DOI:
10.1080/23311975.2023.2275851

To link to this article: https://doi.org/10.1080/23311975.2023.2275851

© 2023 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group.

Published online: 15 Nov 2023.

Submit your article to this journal

Article views: 1347

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=oabm20
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

INFORMATION & TECHNOLOGY MANAGEMENT | REVIEW ARTICLE

ChatGPT: A brief narrative review
Bulbul Gupta1, Tabish Mufti1, Shahab Saquib Sohail1 and Dag Øivind Madsen2*

Abstract: In this study, we present a brief narrative review focused on ChatGPT,

a state-of-the-art conversational agent developed using OpenAI’s Generative
Received: 01 June 2023
Accepted: 23 October 2023 Pretrained Transformer (GPT) framework. Distinctive for its ability to generate text
of high quality in real-time, ChatGPT has emerged as a leader among artificial
*Corresponding author: Dag Øivind
Madsen, School of Business, intelligence chatbots, garnering interest from both commercial and scholarly circles.
University of South-Eastern Norway,
3511 Hønefoss, Norway Our review explores the technological underpinnings of ChatGPT, examines its
E-mail: [email protected] inherent features that support its performance, and analyzes existing research on
Reviewing editor: its applications and impacts across several domains. Through this assessment, we
Balan Sundarakani, University of
Wollongong Faculty of Business,
delineate ChatGPT’s strengths and limitations, offering informed recommendations
United Arab Emirates for future investigations in this burgeoning research field.
Additional information is available at
the end of the article Subjects: Artificial Intelligence; Information & Communication Technology (ICT); Internet
& Multimedia - Computing & IT; Technology

Keywords: ChatGpt; artificial intelligence; generative artificial intelligence; machine

learning; large language models

1. Introduction
Modern technology relies heavily on Artificial Intelligence (AI), which operates covertly to mimic
the human mind and assist us in different ways (Kaplan, 2016). Although AI has a long history,
there have been significant advances in recent years (Haenlein & Kaplan, 2019). These advance­
ments have materialized in the development and launch of AI-powered chatbots such as ChatGPT,
demonstrating to the public how far AI has progressed (Susnjak, 2022).

ChatGPT-3 was developed using an upgraded form of GPT-3, an improved language-developing

AI standard created by OpenAI (Sohail, 2023). The Deep Learning Neural Network (DLNN) utilized in
GPT-3 has almost 175 billion Machine Learning (ML) parameters. To place things in context, the
biggest acquired language model before GPT-3 was Microsoft’s Turing-Natural Language
Generation (T-NLG) framework, which includes 10 billion parameters. By the beginning of 2021,
GPT-3 was the largest Neural Network (NN) ever built. So far as creating content that looks to have
been written by a human, GPT-3 is better than all preceding versions (Khalil & Er, 2023).

The ChatGPT chatbot is built using the OpenAI GPT-3 language structure. It is intended to create
text replies that sound like human answers to operator data entered in a chat setting. With the
help of a vast database of human communications (training data), OpenAI ChatGPT was developed
to provide replies to various subjects and cues. It has been pointed out that Generative AI
language models’ understanding are based on the patterns and structures they learned from
the training data, and these models lack human contextual awareness and understanding
(Bozkurt, 2023). Nevertheless, ChatGPT is able to provide useful responses in many different

© 2023 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution
License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribu­
tion, and reproduction in any medium, provided the original work is properly cited. The terms on
which this article has been published allow the posting of the Accepted Manuscript in
a repository by the author(s) or with their consent.

Page 1 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

languages. Some media articles note that it supports nearly 100 languages, and a research study
evaluated the chatbot’s performance in 37 languages (Lai et al., 2023). Therefore, ChatGPT can be
highly useful for users worldwide, for example, in areas such as language translation, client
support, and content development activities. The OpenAI API makes ChatGPT accessible, allowing
programmers to use and incorporate it into their apps and devices (Gozalo-Brizuela & Garrido-
Merchan, 2023).

ChatGPT has sparked a lot of interest, both in the business world and in academic circles (Farhat
et al., 2023). It has received massive attention in both traditional and social media (Karanouh, 2023;
Leiter et al., 2023). While many supporters have touted the great potential of Generative AI, some
skeptics have noted that the discourse around ChatGPT is dominated by hyperbolic language. Some
strong critics have sounded the alarm bells and argue that AI models such as ChatGPT can pose
a threat to human civilization (Chomsky et al., 2023; Harari, 2023).

Figure 1 shows worldwide search interest for the term “ChatGPT” from 3 March 2022, to
3 March 2023, as measured by Google Trends. The figure shows that the search interest during
this period was highest in China but that the interest was also relatively formidable in several large
countries such as the United States, Canada, India, and Australia. At the same time, search
interest was quite low in countries such as Mexico, Russia, Turkey, Chile, Peru, Argentina, and
Iran. It is also notable that there was very little search activity in many of the African countries.

In recent months, numerous comprehensive and systematic reviews on various aspects of ChatGPT
literature have emerged (e.g., Li et al., 2023; Ray, 2023; Sallam, 2023; Farhat et al., 2023b). Our paper
sets itself apart by utilizing a narrative review approach, which involves an in-depth and critical
examination of the existing literature on a specific subject (Ferrari, 2015). Although a narrative review
may lack the objectivity and rigor of a systematic review, it provides considerable room for subjective
insights and conjectures about future directions in the field of research.

In conducting this narrative literature review, we employed a snowballing type of methodology

that combined both backward and forward snowballing techniques (Felizardo et al., 2016; Wohlin,
2014). For backward snowballing, we traced the references cited in recent ChatGPT papers to
identify important early works. Concurrently, forward snowballing involved tracking citations of
foundational ChatGPT papers to discover more recent studies that build upon or challenge their

Figure 1. ChatGPT worldwide

interest.

Source: Google Trends (trends.

google.com/trends).

Page 2 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

findings. This approach ensured a well-rounded exploration of the existing body of knowledge on
ChatGPT.

In this brief narrative review of the ChatGPT literature, we will explore various facets of this
rapidly developing research area. Topics covered will include the technical foundations of ChatGPT,
its operational mechanisms, strengths, weaknesses, and the factors contributing to its popularity.
Additionally, we will offer insights into potential future trends and research directions.

2. Background of ChatGPT
Silicon Valley has been the epicenter of the development of ChatGPT, and several of the most well-
known business and technology figures have been involved in the development and financing of
chatbot technology. OpenAI, the business that created ChatGPT, was launched as a charity in 2015
by Greg Brockman, Elon Musk, Ilya Sutskever, Wojciech Zaremba, Peter Thiel, and other technology
developers. Its objective was to prevent the centralized control of AI by providing its work openly
to the general population. According to information published on OpenAI’s website on
11 December 2015, the company aims to develop artificial intelligence in a way that is most likely
to benefit humanity (Khan et al., 2023).

Elon Musk resigned from the panel in 2018 because of a conflict of interest with Tesla AI. In
2019, OpenAI changed its status from a non-business entity, to “capped-gain,” which would
enable investors to earn 100× possible profits while still supporting non-profit endeavors with
the leftover funds. In 2019, Microsoft invested $1 billion in OpenAI, and in the last few years, the
company has made further investments in the partnership that allows Microsoft to compete with
Google’s AI business, DeepMind (Lehnert, 2023).

OpenAI has developed ChatGPT over several years. Figure 2 provides a simple overview of the
development process of ChatGPT from GPT-3 to GPT-4. As shown, GPT-3 was introduced in 2020,
and several GPT-3 models have been released. On 30 November 2022, OpenAI released
a downloadable demo of ChatGPT (GPT-3.5). This AI-powered chatbot is able to interact with
human communication and provides answers to queries within a couple of seconds. Within five
days of its release, ChatGPT had already 1 million users. As we indicated in the introduction,
ChatGPT quickly attracted much interest and attention due to its ability to generate thorough and
precise replies to queries across various subject areas. It was the first time such a potent and
accessible chatbot online interface had been freely accessible to the general public. As a result,
OpenAI’s projected valuation increased considerably. Although ChatGPT was presented as a free
service, commentators quickly noted that it was doubtful that the free service would continue to
be available in the future (Deng & Lin, 2022).

Figure 2. ChatGPT development

process.

Page 3 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

In March 2023, another key development in the history of ChatGPT took place. On
14 March 2023, ChatGPT-4 was made public through API and to paying ChatGPT+ subscribers.
The GPT-4 (Generative Pre-Trained Transformer 4) is the 4th version in the GPT series, and it is a big
structure LLM developed by OpenAI. Microsoft acknowledged that earlier iterations of its search
engine Bing that utilized GPT actually did so before GPT-4 was formally released. GPT-4 was taught
to anticipate the coming unit as a transformer implementing both public and private information
and was then enhanced with RL (Reinforcement Learning) via user and AI input for quality
management and human synchronization. The following are some potential improvements that
GPT-4 offers:

● Improved Language Modeling: GPT-4 is anticipated to contain more parameters and to have been
trained on a broader range of data sets, which might result in more accurate and reliable language
modeling skills.
● Multimodal Learning: GPT-4 may be created to learn from a variety of modalities, including text,
graphics, audio, and video, enabling it to comprehend and provide answers across several media types.
● Better Contextual Understanding: The contextual comprehension and reasoning capabilities of GPT-4
may be more sophisticated, enabling it to produce more logical and pertinent replies depending on
the conversation’s context.
● Increased Efficiency: GPT-4 may be quicker and more energy-efficient than its forerunners, opening it
up to a broader variety of applications and devices.
● Enhanced Creativity: Beyond the facts and information it has been educated on, GPT-4 may have
increased creativity and is able to produce more inventive and varied replies.

3. The mechanism supporting ChatGPT

Generative Pre-Training Transformer 3 (GPT-3) is a state-of-the-art AI system. It enables chatbots
to interpret and develop normal language similar to humans with impressive precision and fluency.
It is the broadest language standard created to date with 175 billion parameters and the potential
to quickly action millions of texts (Ufuk, 2023).

A Deep Neural Network (DNN) has already been tested by OpenAI using a sizable sentence
database, and its functionality has been enhanced for purposes like creating sentences or respond­
ing to queries. This is the core tech behind Chat GPT-3. The grid is built from several converter units
that analyze the entered sentence and show results. In addition, the connection has intra-
attention features that allow it to evaluate the significance of various words and terms about
one another and the discussions as a whole. Generators also enable ChatGPT-3 to produce mean­
ingful sentences even from minimal information (Gao et al., 2022).

A noteworthy development in Natural Language Processing (NLP) is ChatGPT-3, which utilizes

a transformer-built structure to analyze massive volumes of information concurrently and create
a language closer to what a human would interpret (Jeblick et al., 2022). This innovation has
several applications, including text categorization services, bots, and automatic translation appli­
cations. Nevertheless, ChatGPT-3 cannot connect to the Web and can only function by utilizing the
Internet it has learned during its development, which restricts its ability to acquire outside knowl­
edge (Rudolph et al., 2023).

Figure 3 displays a word cloud for ChatGPT that is evidence of its vast lexicon and subject-matter
expertise. It displays words from various fields, including technology, science, and current events.
AI-related terms like “Machine Learning, “ChatGPT,” “Neural Networks,” and “Deep Learning” are
included in the word cloud. It also contains words like “natural language processing,” “language
generation,” and “text completion”.

4. Merits of ChatGPT
Since its introduction, there has been an exponential increase in the number of studies on ChatGPT
(Sohail et al., 2023b), some of which have even been co-authored with ChatGPT (e.g., Ali & OpenAI,

Page 4 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

Figure 3. Word cloud of

ChatGPT.

Inc, C., 2023; King & ChatGPT, 2023). Many of the studies have looked at ChatGPT’s impact and
implications in education (Farrokhnia et al., 2023; Zhai, 2022), business (George & George, 2023;
Korzynski et al., 2023) or in the healthcare sector (Li et al., 2023; Sallam, 2023). In the following, we
will briefly review the main findings about the merits of ChatGPT in two of these application areas:
education and healthcare.

4.1. Education
In the field of education, ChatGPT presents substantial benefits. It can be a supplementary tool for
teachers, providing them with resources and content to enhance their teaching methods. For
students, ChatGPT can offer personalized tutoring, help clarify complex concepts, and encourage
self-paced learning (Kasneci et al., 2023). Additionally, it can generate hypothetical scenarios for
various subjects, aiding in practical learning. A comprehensive range of research studies provide
insight into the multiple ways ChatGPT can be integrated into educational systems and the
potential benefits and risks associated with its use, such as cheating and plagiarism (King &
ChatGPT, 2023; Pavlik, 2023; Rudolph et al., 2023; Wang et al., 2022). Studies also show that
ChatGPT performs increasingly well on different entrance and standardized exams and tests across
academic subjects (Gilson et al., 2022; Huh, 2023; Kung et al., 2023; Wood et al., 2023). While there
is still room for development and refinement, the potential of ChatGPT in enhancing educational
experiences is vast.

4.2. Healthcare
Much research has also looked at the merits of ChatGPT in the healthcare sector (Cascella et al.,
2023; Li et al., 2023; Sallam, 2023). ChatGPT can play a significant role in healthcare, mainly by
enhancing accessibility to health information and streamlining health-related processes. For
instance, it can serve as a first point of contact on digital health platforms, providing general
health information, guiding users through symptom checkers, and referring them to appropriate
healthcare resources (Hopkins et al., 2023; Javaid et al., 2023). It can also assist healthcare
professionals by summarizing the latest medical research findings from large databases, aiding
them in staying on top of current developments. However, it is important to be mindful of
ChatGPT’s limits, biases, and risks (Sajjad & Saleem, 2023). Therefore, it should not replace profes­
sional medical advice or consultation, as it may lack the specialized knowledge to provide direct
medical advice.

5. Pros of ChatGPT
Since its launch, ChatGPT has grown in popularity among many demographic groups. However, the
response has been relatively mixed. While some praise ChatGPT for its benefits and future
potential, others remain more skeptical and criticize it for its shortcomings, constraints, and

Page 5 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

possible disadvantages. In the following, we will examine some of the main advantages of
ChatGPT (Srivastava, 2023).

5.1. Ability to mimic human dialogue

The primary role of ChatGPT is to mimic human dialogue based on operator-provided submissions
or commands. It is similar to AI assistant innovations and system apps such as Alexa and Siri. It is
created on more developed Reinforcement Learning (RL) and Supervised Learning (SL) methods
implementing Large Learning models (LLL) algorithm and assessing its functionality, and it imi­
tates real-world discussion.

5.2. Intelligent and adaptable language model

Generative Pre-Trained Transformer-3 (GPT-3) is a decoder and language prophecy structure
designed by OpenAI. It is considered among the most powerful AI methods ever constructed
(Donato et al., 2023). It is tough to decide whether a message is created by an individual due to
the high standard of the messages it creates. Being trained on a sizable collection of text, GPT-3 is
a very intelligent and adaptable language model. Consequently, ChatGPT is versatile enough to
handle multiple tasks due to its extensive data range (Haque et al., 2022).

5.3. Broad-Variety Implementations

ChatGPT can perform multiple functions. It can generate text that compares to that of skilled
Artificial Intelligence (AI) writers. Analyses have shown that it is even proficient in noting songs
and forming imaginary works, such as novels. It can support technical developers or content
supporters in creating a summary (Zhang et al., 2022). Therefore, further exploration of how
ChatGPT can be utilized for such tasks is needed (Pardos & Bhandari, 2023).

5.4. Open to additional improvements

The foundation of ChatGPT is a machine learning model, which can be continually improved by
being trained on fresh data. Through ChatGPT, the knowledge to make improvements in its
responses and available implementation are other benefits. Depending on the presented LLMs,
there is always a chance for development via an effective program utilizing SL and RL. An operator
can offer additional information about whether they like or dislike a specific answer (Hosseini &
Horbach, 2023).

5.5. Natural language understanding

ChatGPT is based on the GPT (Generative Pre-Trained Transformer) architecture, enabling it to
comprehend real language’s syntactic and grammatical structures. It has the ability to detect
typical grammatical constructions and idioms after being trained on a vast corpus of textual data,
which includes books, papers, and websites (Kocoń et al., 2023). This implies that even when the
data it gets is not constructed correctly or includes faults, it may nevertheless provide replies that
are grammatically accurate and semantically relevant (Wang et al., 2023).

5.6. Wide range of applications

Customer support, personal assistance, and content creation are just a few of the uses for
ChatGPT. ChatGPT can assist organizations in automating their customer care assistance proce­
dures, lowering the demand for human agents and enhancing response times. ChatGPT can aid
users with personal assistance chores like making appointments or looking for information online
(Dai et al., 2023). Finally, ChatGPT can create content, such as text for social media postings or
marketing initiatives.

Moreover, ChatGPT is a vital tool for several applications due to its features. For example, it can
be a useful tool for interacting with users and enhancing their experience because of its grasp of
natural language, contextual awareness, and learning capability (Shahriar & Hayawi, 2023).
Moreover, it can be useful for companies and organizations wishing to offer top-notch support or
customer care due to its scalability and around-the-clock availability.

Page 6 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

6. Cons of ChatGPT
The previous section showed that ChatGPT has extensive knowledge of content from various
sources, including books, journals, and web pages. Its ability to recall and provide reliable informa­
tion is critical for several sensitive applications and other essential AI technologies. However, it is
not perfect, and its accuracy can be compromised, as it relies on a learning algorithm. Sometimes,
ChatGPT may provide biased or fabricated information.

Therefore, ChatGTP is not without its shortcomings and struggles with some of the same
problems as many other chatbots in the past. In the following, we will discuss some of the primary
drawbacks identified in previous studies (Kasneci et al., 2023).

6.1. Lack of clarity and factual errors

The point that ChatGPT periodically can develop sentences that appear precise or effective but are
incorrect or illogical is among the main faults and shortcomings (Wang et al., 2023). Sometimes,
ChatGPT cannot completely comprehend a question due to a lack of context, which might result in
confused or inaccurate answers. For instance, according to several studies (Farhat et al., 2023;
Sohail, 2023; Sohail et al., 2023; Wang et al., 2023), if a user poses a question that depends on
details from an earlier exchange, ChatGPT might not be aware of that context and might give an
answer that is inaccurate or ambiguous. It is systematic in statistical language standards and is
known as “commotion” and highlights the challenge of contextual comprehension in AI language
models. Moreover, ChatGPT provides no sources or footnotes regarding where to discover the
content. Therefore, it is not optimal to implement this chatbot by itself for digital tracking and
study (Kuzman et al., 2023).

6.2. Poor understanding of recent developments

The version of ChatGPT introduced in November 2022 is limited to providing information on events
that occurred up to 2021. As it continues to generate content based on text created by humans, it
will eventually include references to more current events (Jiao et al., 2023). A notable downside of
ChatGPT is its limited understanding of current events. This occurs because the system’s training is
based on a static text dataset, which may not include the most recent facts or developments (Cao
et al., 2023). However, the recent developments related to web browsing in ChatGPT-4 may
remedy this problem. Moreover, the introduction of extensions such as WebChatGPT, a web-
based version of the technology, appears to reduce some of the errors by directing the queries
to specific databases or sites.

6.3. Problems and questions of ethics

The use of ChatGPT also raised numerous ethical issues (Rahimi & Abadi, 2023; Zhuo et al.,
2023). Numerous universities and schools have considered limiting access to ChatGPT or
banning its use entirely (Ahmad et al., 2023; Hsu, 2023). Because its results rely on human-
created sentences, academics and creators have worried about copyright violations (Cooper,
2023). Unintentionally spreading false information or fake news with ChatGPT might have
negative repercussions, and this may occur if ChatGPT is not educated on trustworthy informa­
tion sources or accuracy is not prioritized above interaction. This also raises questions about
the appropriateness of using it in operations that require human association, including 24/7
help or psychological counseling (Antaki et al., 2023).

6.4. Possible legal issues

ChatGPT was developed using data from The Common Crawl database, which includes copyrighted
content from publishers and works by individual authors and scholars (Sinha et al., 2023).
Consequently, there is a possibility that guidance dispensed by ChatGPT in legal or financial
matters may not be accurate or fully up-to-date. Individuals or businesses acting on such advice
could find themselves liable for any resulting issues. Experts have also cautioned against the risk of
using AI-generated services for illicit activities, including computer crime.

Page 7 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

6.5. Limited domain expertise

As a general-purpose language model, ChatGPT has certain limitations when it comes to specia­
lized domains (McGee, 2023). Although it is capable of generating responses that are grammati­
cally correct and semantically coherent, it may lack the specialized expertise required to provide
accurate responses to queries in specific fields. For example, when confronted with a user asking
for a medical consultation, ChatGPT might fall short of providing an appropriate response
(Halaweh, 2023). This is because it lacks the domain-specific medical knowledge that a qualified
healthcare professional would possess. Despite its extensive training data, it cannot replace expert
advice in fields that require specialized education and training.

6.6. Biased responses

Large volumes of text data, some of which may be skewed or problematic, are used to train
ChatGPT. This means it may create replies containing prejudice or preconceptions, particularly if
the training data have such biases (Xue et al., 2023). If ChatGPT was trained using biased data, it
would pick up on such biases and may repeat them in its answers. For instance, ChatGPT may yield
results that reflect prejudices or discrimination if the training data contains such biases. This is
a well-known problem with many AI models, and work is being done to fix it by using more
representative and varied training data (AlAfnan et al., 2023).

ChatGPT’s drawbacks are attributed to its difficulties in comprehending specialized topics,

propensity for producing biased replies, lack of emotional intelligence, and restricted compre­
hension of context. However, even though these drawbacks may be considerable, continuous
work in the field of natural language processing is being done to find solutions (Zhu et al.,
2023).

7. The future of ChatGPT

Currently, ChatGPT technology has gained traction in many areas of business and society (George
& George, 2023; Haleem et al., 2023; Hassani & Silva, 2023). Figure 4 provides an overview and
illustration of future applications of ChatGPT in different sectors of the economy. As we move
forward, there is potential for this technology to be refined and streamlined, potentially revolutio­
nizing the way humans create value. However, alongside these advancements, concerns are
emerging about the technology’s potential impact on employment, stirring long-standing debates
about how AI might transform or even displace certain job roles (Bessen, 2018; Davenport &
Westerman, 2018).

In the following, we will look at a few of ChatGPT’s applications and speculate about ChatGPT’s
future development and trajectory. Moreover, we will develop several proposition and indicate how
these can be tested in future research.

Figure 4. Overview of future

applications of ChatGPT in dif­
ferent sectors.

Page 8 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

7.1. Customer support

ChatGPT is also likely to affect how businesses interact and engage with consumers (Dwivedi et al.,
2023; Paul et al., 2023). AI-based chatbots may respond to consumer inquiries quickly and provide
round-the-clock client service. The enthusiasm that ChatGPT has generated is evidence of its
ingenuity and growing significance in the AI industry. There is significant scope for ChatGPT to
be connected with current interactive AI to enhance user support discussions as the technology
develops, and many firms use AI innovations to enhance the customer experience. In the future,
by responding to customers’ wants and questions in a quicker and more personalized way, they
may be able to dramatically enhance the level of customer service (Sanmarchi et al., 2023).

Proposition: The integration of ChatGPT-powered chatbots in customer support processes will lead
to a reduction in average response time and an improvement in customer satisfaction
scores. To test this proposition, researchers could conduct experiments or collect data
from businesses that have implemented ChatGPT in their customer support opera­
tions. They can compare response times and customer satisfaction scores before and
after ChatGPT integration to determine if there is a difference, providing valuable
insights into the impact of ChatGPT on customer support efficiency and customer
satisfaction.

7.2. Marketing and sales

ChatGPT has a multitude of applications in marketing and sales (Dwivedi et al., 2023; George &
George, 2023; Rivas & Zhao, 2023). By helping clients in the purchasing process, chatbots can aid in
information processing, outlook assessment, and proposed settlement. A catchy title is necessary
for written material to do effectively on search engines, making this a crucial advantage of creative
AI technologies like ChatGPT. Marketing professionals may also use this AI tool’s power and ability
to create intriguing headlines and content for articles, seminars, and conferences. Employers can
use chatbots to simplify several steps in the hiring and recruitment process, boosting productivity
and reducing costs and time (Korzynski et al., 2023). AI-powered chatbots can also assist employ­
ers in simplifying the hiring process by pre-screening applicants. Moreover, in the future, chatbots
can likely compile key information about possible clients and offer parameterized testing recom­
mendations to the marketing and sales teams, allowing them to customize their strategy.

Proposition: The integration of ChatGPT-powered chatbots in marketing and sales processes will
lead to an increase in lead conversion rates and a reduction in customer acquisition
costs. To test this proposition, researchers could analyze marketing and sales data
from businesses that have integrated ChatGPT-powered chatbots into their processes.
They can compare lead conversion rates and customer acquisition costs before and
after ChatGPT implementation to determine if there is a difference. Such studies could
shed light on the effectiveness of ChatGPT in enhancing marketing and sales out­
comes.

7.3. Education
AI-based chatbots like ChatGPT have many applications in education (Baidoo-Anu & Owusu Ansah,
2023; Bozkurt, 2023; Bozkurt et al., 2023; Farrokhnia et al., 2023; Kasneci et al., 2023; Tlili et al.,
2023) and have emerged as a useful tool for instructors to educate students and provide them
with engaging and informative responses to their questions. Gill et al. (2023) argue that ChatGPT
can have “transformative effects” on the learning environment. Among the educational applica­
tions of ChatGPT, one possibility is the creation of quizzes and tests that can help evaluate the
learners’ knowledge and progress. ChatGPT is known for its advanced capabilities and extensive
data resources, which make it a powerful educational tool. However, it is essential to note that the
dataset used to train ChatGPT is compiled from various online sources, some of which may contain

Page 9 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

errors or inaccuracies. For example, several authors have noted that ChatGPT sometimes makes up
facts and “hallucinates” (Alkaissi & McFarlane, 2023; Azamfirei et al., 2023), for example, by
generating “ghost” bibliographic references that sound plausible but in reality are fake
(Orduña-Malea & Cabezas-Clavijo, 2023). This indicates that students should utilize ChatGPT as
a reference comparable to Wikipedia, which has also been criticized for inaccuracies (Holman
Rector, 2008).

Therefore, it is crucial for instructors to carefully review and validate the responses provided by
ChatGPT before incorporating them into their instructional materials. However, with the right
precautions and monitoring, ChatGPT can be an effective tool for enhancing students’ learning
experience. In particular, ChatGPT can be a useful approach for a broad understanding and
a preliminary step in the learning process (Cotton et al., 2023).

Proposition: The integration of ChatGPT-powered virtual tutors in educational settings will lead to
an improvement in student engagement, knowledge retention, and learning outcomes.
To test this proposition, researchers could carry out controlled experiments or long­
itudinal studies in educational settings where ChatGPT is used as a supplementary
teaching tool. They can measure student engagement levels, knowledge retention,
and learning outcomes, comparing the results of groups with and without ChatGPT
assistance. This research can provide insights into the potential benefits of using
ChatGPT in education.

7.4. Healthcare
Much research has looked at the prospects of ChatGPT in the healthcare context (Cascella et al.,
2023; Javaid et al., 2023; Li et al., 2023; Sallam, 2023; Vaishya et al., 2023). AI-based chatbots can
provide individualized health advice and assistance in identifying medical issues. Patients may find
it difficult to remember to take their prescriptions on schedule and to adhere to their physician’s
dose recommendations, especially if they are taking many medications. Patients can utilize
ChatGPT to handle their medication regimens, containing alerts, dose guidelines, and possible
bad impacts. Medicine use, restrictions, and other vital factors impacting pharmacological inter­
vention can also be discussed with patients via ChatGPT. Furthermore, ChatGPT can help patients
have a better healthcare experience by automating organizational activities like consultation
registration and treatment plan updates (Jeblick et al., 2022). However, it is important to keep in
mind the chatbot’s limitations since research has found that it sometimes gives erroneous
responses to medical questions (Vaishya et al., 2023).

Proposition: The integration of ChatGPT-powered virtual healthcare assistants into clinical work­
flows will lead to a reduction in healthcare administrative burden, improved patient
engagement, and increased overall healthcare efficiency. To test this proposition,
researchers could conduct studies in healthcare settings where ChatGPT is used to
assist healthcare professionals and interact with patients. Specifically, researchers
could measure changes in administrative workloads, patient engagement levels, and
overall healthcare efficiency, comparing the results of healthcare facilities that have
integrated ChatGPT with those that have not. This research could yield insights into
the potential benefits of using ChatGPT in healthcare.

7.5. Language translation

ChatGPT can be utilized for language translation problems since it has a comprehensive language
model (Jiao et al., 2023). It can be trained on huge datasets of parallel text and text in two
languages that mean the same thing. It can then be implemented to translate text from one
language to another because of its capacity to produce a coherent text. By learning the

Page 10 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

connections between vocabulary and grammatical conventions in both languages, ChatGPT, for
instance, may produce a similar phrase in one language given a sentence in another language. It
can generate appropriate translations by considering a sentence’s context and meaning
(Zarifhonarvar, 2023).

Proposition: The utilization of ChatGPT for language translation will lead to improvements in the
accuracy and efficiency of cross-language communication, making it a valuable tool for
bridging language barriers. To test this proposition, researchers could conduct experi­
ments or evaluations comparing the accuracy and efficiency of ChatGPT-powered
language translation with traditional translation methods or other machine transla­
tion systems. More specifically, researchers could collect data on translation quality,
speed, and user satisfaction, comparing the results between the use of ChatGPT and
other methods. Such research could yield valuable insights into the effectiveness of
ChatGPT in improving language translation processes.

7.6. Travel
ChatGPT also has a multitude of applications in travel, tourism, and hospitality (Ali & OpenAI, Inc,
C., 2023; Carvalho & Ivanov, 2023; Dwivedi et al., 2023; Gursoy et al., 2023; Iskender, 2023; Ivanov
& Soliman, 2023). Chatbots have become increasingly popular in the tourism industry, providing
visitors with a seamless and convenient way to book tickets, hotel rooms, and other accommoda­
tions. Additionally, AI-powered chatbots have the potential to revolutionize the tourism industry by
offering real-time information on climate conditions, local events, and flight details, making it
easier for travelers to plan their journeys and stay up to date on any changes or delays. Altogether,
ChatGPT can revolutionize the tourism and hospitality sector by arranging trips in a better way,
enhancing the guest experience, offering language translation services, enhancing promotional
activities, and allowing for advanced analytics (Ali & OpenAI, Inc, C., 2023).

Proposition: The integration of ChatGPT-powered virtual travel assistants in travel booking and
planning services will lead to an increase in customer satisfaction, more personalized
travel experiences, and improved operational efficiency for travel companies. To test
this proposition, researchers could conduct studies or surveys comparing the experi­
ences and satisfaction levels of travelers who have used ChatGPT-powered virtual
travel assistants with those who have not. More specifically, researchers could collect
data on customer satisfaction, personalized travel recommendations, and opera­
tional efficiency improvements within travel companies. This type of research would
provide insights into the potential benefits of using ChatGPT in the travel industry.

7.7. Entertainment
Creative writing is one area where ChatGPT does well (Taecharungroj, 2023), and therefore, it
has various promising applications in the entertainment industry. For instance, AI-powered
chatbots could offer personalized recommendations for TV shows, music albums, and various
other forms of entertainment tailored to individual tastes and preferences. Additionally, they
have the potential to generate complete scripts for movies or TV shows and even craft lyrics for
songs. The possibility of AI in reshaping and enhancing the entertainment experience is vast
and intriguing. Although some are skeptical (Bhandari, 2023), it is likely that ChatGPT will
become a helpful technique for the film industry, providing various benefits, such as audience
commentary and actionable insights into directing, writing, and personality creation. In this
manner, directors may use AI to generate more interesting and popular movies while preser­
ving the craftsmanship and innovation that have long been at the core of the movie industry
(Haensch et al., 2023).

Page 11 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

Proposition: The integration of ChatGPT-powered virtual entertainment assistants in streaming plat­

forms and gaming will lead to an increase in user engagement, personalized content
recommendations, and overall entertainment satisfaction. To test this proposition,
researchers could conduct user studies or surveys comparing the engagement levels,
content recommendations, and satisfaction of users who have interacted with ChatGPT-
powered virtual entertainment assistants with those who have not. They could collect
data on user preferences, content consumption, and user feedback to evaluate the
impact of ChatGPT on entertainment experiences. This type of research would shed
light on the potential benefits of using ChatGPT in the entertainment industry.

7.8. Programming
The emergence of AI programmers like ChatGPT will also result in a growth in the need for computer
programmers knowledgeable in data science methods. For instance, developers proficient in Go,
Python, and other advanced analytics tools and languages can develop, implement, and deploy
applications. Moreover, programmers can write programs faster, more easily, and quickly by using
ChatGPT, which can produce program clips depending on particular computer programming and
patterns. Merow et al. (2023) found that ChatGPT could be highly useful for speeding up monotonous
parts of coding, which allows users to free up time and energy to focus on the more challenging parts.
Finally, ChatGPT can aid in the review process and troubleshooting (Sobania et al., 2023).

Proposition: The integration of ChatGPT-powered code assistants in programming environments will

lead to an improvement in developer productivity, code quality, and knowledge transfer
among software development teams. To test this proposition, researchers could con­
duct experiments or case studies involving software development teams using
ChatGPT-powered code assistants and compare their productivity, code quality, and
knowledge transfer experiences with those who do not have access to such tools.
Data on code development metrics, error rates, and team dynamics can be collected
and analyzed to evaluate the impact of ChatGPT on programming-related tasks and
collaboration within development teams. This type of research would be able to cast
light on the potential benefits of using ChatGPT in the field of programming.

7.9. Personal assistants

With its capacity to interpret natural language and produce coherent text, ChatGPT has demon­
strated considerable promise as a personal assistant providing 24/7 communication, for example,
for tourists (Wong et al., 2023). To give consumers even more in-depth support, in the future, it
could be integrated with other software programs, such as email and calendar apps. To increase its
precision and effectiveness over time, ChatGPT may employ machine learning techniques. It will
learn from user input and interactions to better comprehend specific preferences and require­
ments. Therefore, ChatGPT may become a more potent tool for productivity and personal organiz­
ing as technology develops and natural language processing improves (Murk et al., 2023).

Proposition: The integration of ChatGPT as a personal assistant will lead to an improvement in individual
productivity, organization, and information retrieval, resulting in enhanced personal and
professional effectiveness. To test this proposition, researchers could conduct user studies
or surveys comparing the productivity levels, organization, and information retrieval
experiences of individuals who use ChatGPT as a personal assistant with those who do
not. For example, researchers could collect data on task completion times, task manage­
ment efficiency, and user satisfaction to evaluate the impact of ChatGPT on personal and
professional effectiveness. Such research would be able to shed light on the potential
benefits of using ChatGPT as a personal assistant.

Page 12 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

8. Conclusion
According to this brief narrative review of ChatGPT, it is undeniable that AI technology has
advanced significantly, and AI language tools like ChatGPT have a multitude of real-world applica­
tions. Furthermore, it is evident that ChatGPT stands out as a robust natural language processing
system capable of generating text that closely mimics human-like speech and writing. By employ­
ing advanced AI technologies like ChatGPT, businesses and society at large can reap a host of
benefits. These include increased operational efficiency, heightened accuracy in information deliv­
ery, and substantial cost reductions. These advantages make such technologies increasingly
integral in a wide range of applications, from customer service to content generation. That said,
it is important to consider the limitations, which include security issues and restricted capabilities.
Despite these challenges, ChatGPT is a rapidly advancing AI platform capable of automating
conversations and delivering more accurate responses. At the same time, it is important to note
that human oversight is still required to ensure the generated content meets specific standards.
Therefore, it cannot, for the time being, fully substitute for human authors.

The primary objective of this paper was to conduct a brief narrative review of ChatGPT, an AI-
powered conversational chatbot. Given the nature and scope of a narrative review, as opposed to
a systematic review or meta-analysis, there are several limitations that readers should keep in mind.

Firstly, the brevity of the review implies that not all aspects of ChatGPT could be covered
exhaustively. The paper provides a high-level overview of ChatGPT research rather than an in-
depth analysis, and as a result, some nuances or complexities might have been overlooked.
Secondly, narrative reviews are inherently prone to author bias, as they often do not follow
a strict methodological framework for selecting and evaluating the literature. Consequently,
this review may only capture part of the spectrum of perspectives or findings related to
ChatGPT. Thirdly, the rapidly evolving landscape of AI and conversational agents means that
the information presented might quickly become outdated. Future versions of ChatGPT (e.g.
GPT-5) or competing technologies such as Google Bard could offer features and capabilities not
covered in this review, making the findings less applicable over time. Fourthly, the paper does
not provide quantitative analysis or statistical evaluations, which would be more rigorous in
assessing the merits or limitations of ChatGPT. This leaves our critique largely qualitative and
interpretive in nature. Lastly, the review does not encompass all available research or docu­
mentation related to ChatGPT, including proprietary studies, which could offer further insights
into its capabilities and limitations. Therefore, while this narrative review aims to provide
a foundational understanding of ChatGPT, it is important to view its findings within the context
of these limitations.

Author details References

Bulbul Gupta1 Ahmad, N., Murugesan, S., & Kshetri, N. (2023). Generative
Tabish Mufti1 Artificial intelligence and the education sector.
Shahab Saquib Sohail1 Computer, 56(6), 72–76. https://doi.org/10.1109/MC.
ORCID ID: http://orcid.org/0000-0002-5944-7371 2023.3263576
Dag Øivind Madsen2 AlAfnan, M. A., Dishari, S., Jovic, M., & Lomidze, K. (2023).
E-mail: [email protected] ChatGPT as an educational tool: Opportunities, chal­
ORCID ID: http://orcid.org/0000-0001-8735-3332 lenges, and recommendations for communication,
1
Department of Computer Science and Engineering, SEST, business writing, and composition courses. Journal of
Jamia Hamdard, New Delhi, India. Artificial Intelligence and Technology. https://doi.org/
2
School of Business, University of South-Eastern Norway, 10.37965/jait.2023.0184
Hønefoss, Norway. Ali, F., & OpenAI, Inc, C. (2023). Let the devil speak for
itself: Should ChatGPT be allowed or banned in hos­
Disclosure statement pitality and tourism schools? Journal of Global
No potential conflict of interest was reported by the Hospitality and Tourism, 2(1), 1–6. https://doi.org/10.
author(s). 5038/2771-5957.2.1.1016
Alkaissi, H., & McFarlane, S. I. (2023). Artificial halluci­
Citation information nations in ChatGPT: Implications in scientific
Cite this article as: ChatGPT: A brief narrative review, writing. Cureus, 15(2). https://doi.org/10.7759/cur
Bulbul Gupta, Tabish Mufti, Shahab Saquib Sohail & Dag eus.35179
Øivind Madsen, Cogent Business & Management (2023), Antaki, F., Touma, S., Milad, D., El-Khoury, J., & Duval, R.
10: 2275851. (2023). Evaluating the performance of chatgpt in

Page 13 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

ophthalmology: An analysis of its successes and Donato, H., Escada, P., & Villanueva, T. (2023).
shortcomings. medRxiv: 2023. 2001. 2022.23284882. A Transparência da Ciência com o ChatGPT e as
Azamfirei, R., Kudchadkar, S. R., & Fackler, J. (2023). Large Ferramentas Emergentes de Inteligência Artificial:
language models and the perils of their Como se Devem Posicionar as Revistas Científicas
hallucinations. Critical Care, 27(1), 1–2. https://doi. Médicas? The Transparency of Science with ChatGpt
org/10.1186/s13054-023-04393-x and the Emerging Artificial Intelligence Language
Baidoo-Anu, D., & Owusu Ansah, L. (2023). Education in Models: Where Should Medical Journals Stand?.
the era of generative artificial intelligence (AI): https://doi.org/10.20344/amp.19694
Understanding the potential benefits of ChatGPT in Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L.,
promoting teaching and learning. SSRN Electronic Jeyaraj, A., Kar, A. K., Baabdullah, A. M.,
Journal. Available at SSRN 4337484. https://doi.org/ Koohang, A., Raghavan, V., Ahuja, M., Albanna, H.,
10.2139/ssrn.4337484 Albashrawi, M. A., Al-Busaidi, A. S.,
Bessen, J. (2018). AI and jobs: The role of demand. Balakrishnan, J., Barlette, Y., Basu, S., Bose, I.,
National Bureau of Economic Research. Brooks, L. . . . Wirtz, J. (2023). Opinion paper: “so
Bhandari, K. S. (2023). How ChatGPT could harm the film what if ChatGPT wrote it?” multidisciplinary per­
industry. Entrepreneur India. https://www.entrepre spectives on opportunities, challenges and impli­
neur.com/en-in/news-and-trends/how-chat-gpt- cations of generative conversational AI for
could-harm-the-film-industry/444494 research, practice and policy. International Journal
Bozkurt, A. (2023). Generative artificial intelligence (AI) of Information Management, 71, 102642. https://
powered conversational educational agents: The doi.org/10.1016/j.ijinfomgt.2023.102642
inevitable paradigm shift. Asian Journal of Distance Farhat, F., Silva, E. S., Hassani, H., Madsen, D. Ø.,
Education, 18(1). https://www.asianjde.com/ojs/ Sohail, S. S., Himeur, Y., Alam, M. A., & Zafar, A.
index.php/AsianJDE/article/view/718 (2023). Analyzing the scholarly footprint of ChatGPT:
Bozkurt, A., Xiao, J., Lambert, S., Pazurek, A., Crompton, H., Mapping the progress and identifying future trends.
Koseoglu, S., Farrow, R., Bond, M., Nerantzi, C., Preprints.org doi:https://doi.org/10.20944/pre
Honeychurch, S., Bali, M., Dron, J., Mir, K., Stewart, B., prints202306.2100.v1.
Costello, E., Mason, J., Stracke, C., Romero-Hall, E., Farhat, F., Sohail, S. S., & Madsen, D. Ø. (2023). How
Koutropoulos, A., & Jandrić, P. (2023). Speculative trustworthy is ChatGPT? The case of bibliometric
futures on ChatGPT and generative artificial intelli­ analyses. Cogent Engineering, 10(1), 2222988.
gence (AI): A collective reflection from the educa­ https://doi.org/10.1080/23311916.2023.2222988
tional landscape. Asian Journal of Distance Education, Farrokhnia, M., Banihashem, S. K., Noroozi, O., & Wals, A.
18(1), 53–130. https://www.asianjde.com/ojs/index. (2023). A SWOT analysis of ChatGPT: Implications for
php/AsianJDE/article/view/709 educational practice and research. Innovations in
Cao, Y., Li, S., Liu, Y., Yan, Z., Dai, Y., Yu, P. S., & Sun, L. Education and Teaching International, 1–15. https://
(2023). A comprehensive survey of AI-Generated doi.org/10.1080/14703297.2023.2195846
content (AIGC): A history of Generative AI from GAN to Felizardo, K. R., Mendes, E., Kalinowski, M., Souza, É. F., &
ChatGPT. arXiv preprint arXiv:2303.04226. Vijaykumar, N. L. 2016. Using forward snowballing to
Carvalho, I., & Ivanov, S. (2023). ChatGPT for tourism: update systematic reviews in software engineering.
Applications, benefits and risks. Tourism Review. Paper presented at the Proceedings of the 10th ACM/
https://doi.org/10.1108/TR-02-2023-0088 IEEE International Symposium on Empirical Software
Cascella, M., Montomoli, J., Bellini, V., & Bignami, E. (2023). Engineering and Measurement (pp. 1–6). https://doi.
Evaluating the feasibility of ChatGPT in healthcare: org/10.1145/2961111.2962630.
An analysis of multiple clinical and research Ferrari, R. (2015). Writing narrative style literature
scenarios. Journal of Medical Systems, 47(1), 33. reviews. Medical Writing, 24(4), 230–235. https://doi.
https://doi.org/10.1007/s10916-023-01925-4 org/10.1179/2047480615Z.000000000329
Chomsky, N., Roberts, I., & Watumull, J. 2023. Noam Gao, C. A., Howard, F. M., Markov, N. S., Dyer, E. C.,
Chomsky: The false promise of ChatGPT. The Ramesh, S., Luo, Y., & Pearson, A. T. (2022).
New York Times, 8. Comparing scientific abstracts generated by
Cooper, G. (2023). Examining science education in ChatGPT to original abstracts using an artificial
ChatGPT: An exploratory study of generative artificial intelligence output detector, plagiarism detector,
intelligence. Journal of Science Education and and blinded human reviewers. bioRxiv: 2022.2012.
Technology, 32(3), 444–452. https://doi.org/10.1007/ 2023.521610.
s10956-023-10039-y George, A. S., & George, A. H. (2023). A review of ChatGPT
Cotton, D. R., Cotton, P. A., & Shipway, J. R. (2023). AI’s impact on several business sectors. Partners
Chatting and cheating: Ensuring academic integrity Universal International Innovation Journal, 1(1), 9–23.
in the era of ChatGPT. Innovations in Education and Gill, S. S., Xu, M., Patros, P., Wu, H., Kaur, R., Kaur, K.,
Teaching International, 1–12. https://doi.org/10.1080/ Fuller, S., Singh, M., Arora, P., Parlikad, A. K.,
14703297.2023.2190148 Stankovski, V., Abraham, A., Ghosh, S. K.,
Dai, H., Liu, Z., Liao, W., Huang, X., Wu, Z., Zhao, L., Liu, W., Lutfiyya, H., Kanhere, S. S., Bahsoon, R., Rana, O.,
Liu, N., Li, S., & Zhu, D. (2023). ChatAug: Leveraging Dustdar, S. . . . Buyya, R. (2023). Transformative
ChatGPT for text data augmentation. arXiv preprint effects of ChatGPT on modern education: Emerging
arXiv:2302.13007. era of AI chatbots. Internet of Things and Cyber-
Davenport, T. H., & Westerman, G. (2018). Why so many Physical Systems, 4, 19–23. https://doi.org/10.1016/
high-profile digital transformations fail. Harvard j.iotcps.2023.06.002
Business Review, 9. https://hbr.org/2018/03/why-so- Gilson, A., Safranek, C., Huang, T., Socrates, V., Chi, L.,
many-high-profile-digital-transformations-fail Taylor, R. A., & Chartash, D. (2022). How well does
Deng, J., & Lin, Y. (2022). The benefits and challenges of ChatGPT do when taking the medical licensing exams?
ChatGPT: An overview. Frontiers in Computing and The implications of large language models for medical
Intelligent Systems, 2(2), 81–83. https://doi.org/10. education and knowledge assessment. medRxiv:
54097/fcis.v2i2.4465 2022.2012. 2023.22283901.

Page 14 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

Gozalo-Brizuela, R., & Garrido-Merchan, E. C. (2023). 34, 3414–3414. https://doi.org/10.54055/ejtr.v34i.

ChatGPT is not all you need. A state of the art review 3169
of large Generative AI models. arXiv preprint Ivanov, S., & Soliman, M. (2023). Game of algorithms:
arXiv:2301.04655. ChatGPT implications for the future of tourism edu­
Gursoy, D., Li, Y., & Song, H. (2023). ChatGPT and the cation and research. Journal of Tourism Futures, 9(2),
hospitality and tourism industry: An overview of 214–221. ahead-of-print(ahead-of-print). https://doi.
current trends and future research directions. org/10.1108/JTF-02-2023-0038
Journal of Hospitality Marketing & Management, 32 Javaid, M., Haleem, A., & Singh, R. P. (2023). ChatGPT for
(5), 579–592. https://doi.org/10.1080/19368623. healthcare services: An emerging stage for an inno­
2023.2211993 vative perspective. BenchCouncil Transactions on
Haenlein, M., & Kaplan, A. (2019). A brief history of artifi­ Benchmarks, Standards and Evaluations, 3(1),
cial intelligence: On the past, present, and future of 100105. https://doi.org/10.1016/j.tbench.2023.
artificial intelligence. California Management Review, 100105
61(4), 5–14. https://doi.org/10.1177/ Jeblick, K., Schachtner, B., Dexl, J., Mittermeier, A.,
0008125619864925 Stüber, A. T., Topalis, J., Weber, T., Wesp, P., Sabel, B.,
Haensch, A.-C., Ball, S., Herklotz, M., & Kreuter, F. (2023). & Ricke, J. 2022. ChatGPT makes Medicine easy to
Seeing ChatGPT through students’ eyes: An analysis of swallow: An exploratory case study on simplified
TikTok data. arXiv preprint arXiv:2303.05349. radiology reports. arXiv preprint arXiv:2212.14882.
Halaweh, M. (2023). ChatGPT in education: Strategies for Jiao, W., Wang, W., Huang, J.-T., Wang, X., & Tu, Z. (2023).
responsible implementation. Contemporary Is ChatGPT a good translator? A preliminary study.
Educational Technology, 15(2), ep421. https://doi.org/ arXiv preprint arXiv:2301.08745.
10.30935/cedtech/13036 Kaplan, J. (2016). Artificial intelligence: What everyone
Haleem, A., Javaid, M., & Singh, R. P. (2023). An era of needs to know. Oxford University Press.
ChatGPT as a significant futuristic support tool: Karanouh, M. (2023). Mapping ChatGPT in mainstream
A study on features, abilities, and challenges. media: Early quantitative insights through Sentiment
BenchCouncil Transactions on Benchmarks, Standards analysis and word frequency analysis. arXiv preprint
and Evaluations, 2(4), 100089. https://doi.org/10. arXiv:2305.18340.
1016/j.tbench.2023.100089 Kasneci, E., Seßler, K., Küchemann, S., Bannert, M.,
Haque, M. U., Dharmadasa, I., Sworna, Z. T., Dementieva, D., Fischer, F., Gasser, U., Groh, G.,
Rajapakse, R. N., & Ahmad, H. (2022). “I think this is Günnemann, S., Hüllermeier, E., Krusche, S.,
the most disruptive technology”: Exploring sentiments Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J.,
of ChatGPT early adopters using Twitter data. arXiv Poquet, O., Sailer, M., Schmidt, A. . . . Kuhn, J. (2023).
preprint arXiv:2212.05856. ChatGPT for good? On opportunities and challenges
Harari, Y. N. (2023). Yuval Noah Harari argues that AI has of large language models for education. Learning and
hacked the operating system of human civilisation. Individual Differences, 103, 102274. https://doi.org/
The Economist. https://www.economist.com/by-invi 10.1016/j.lindif.2023.102274
tation/2023/04/28/yuval-noah-harari-argues-that-ai- Khalil, M., & Er, E. (2023). Will ChatGPT get you caught?
has-hacked-the-operating-system-of-human- Rethinking of plagiarism detection. arXiv preprint
civilisation arXiv:2302.04335.
Hassani, H., & Silva, E. S. (2023). The role of ChatGPT in Khan, R. A., Jawaid, M., Khan, A. R., & Sajjad, M. (2023).
data science: How AI-Assisted conversational inter­ ChatGPT-Reshaping medical education and clinical
faces are revolutionizing the field. Big Data and management. Pakistan Journal of Medical Sciences,
Cognitive Computing, 7(2), 62. https://doi.org/10. 39(2). https://doi.org/10.12669/pjms.39.2.7653
3390/bdcc7020062 King, M. R., & ChatGPT. (2023). A conversation on arti­
Holman Rector, L. (2008). Comparison of Wikipedia and ficial intelligence, chatbots, and plagiarism in
other encyclopedias for accuracy, breadth, and depth higher education. Cellular and Molecular
in historical articles. Reference Services Review, 36(1), Bioengineering, 16(1), 1–2. https://doi.org/10.1007/
7–22. https://doi.org/10.1108/00907320810851998 s12195-022-00754-8
Hopkins, A. M., Logan, J. M., Kichenadasse, G., & Kocoń, J., Cichecki, I., Kaszyca, O., Kochanek, M.,
Sorich, M. J. (2023). Artificial intelligence chatbots will Szydło, D., Baran, J., Bielaniewicz, J., Gruza, M.,
revolutionize how cancer patients access informa­ Janz, A., Kanclerz, K., Kocoń, A., Koptyra, B.,
tion: ChatGPT represents a paradigm-shift. JNCI Mieleszczenko-Kowszewicz, W., Miłkowski, P.,
Cancer Spectrum, 7(2). https://doi.org/10.1093/jncics/ Oleksy, M., Piasecki, M., Radliński, Ł., Wojtasik, K.,
pkad010 Woźniak, S., & Kazienko, P. (2023). ChatGPT: Jack of
Hosseini, M., & Horbach, S. P. (2023). Fighting reviewer all trades, master of none. Information Fusion 99,
fatigue or amplifying bias? Considerations and 101861. arXiv preprint arXiv:2302.10724. https://doi.
recommendations for use of ChatGPT and other large org/10.1016/j.inffus.2023.101861
language models in scholarly peer review. Research Korzynski, P., Mazurek, G., Altmann, A., Ejdys, J.,
Integrity and Peer Review, 8(1). https://doi.org/10. Kazlauskaite, R., Paliszkiewicz, J., Wach, K., &
1186/s41073-023-00133-5 Ziemba, E. (2023). Generative artificial intelligence as
Hsu, J. (2023). Should schools ban AI chatbots? New a new context for management theories: Analysis of
Scientist, 257(3422), 15. https://doi.org/10.1016/ ChatGPT. Central European Management Journal,
S0262-4079(23)00099-4 ahead-of-print(ahead-of-print), 31(1), 3–13. https://
Huh, S. (2023). Issues in the 3rd year of the COVID-19 doi.org/10.1108/CEMJ-02-2023-0091
pandemic, including computer-based testing, study Kung, T. H., Cheatham, M., Medenilla, A., Sillos, C., De
design, ChatGPT, journal metrics, and appreciation to Leon, L., Elepaño, C., Madriaga, M., Aggabao, R., Diaz-
reviewers. Journal of Educational Evaluation for Candido, G., Maningo, J., Tseng, V., & Dagan, A.
Health Professions, 20, 5. https://doi.org/10.3352/ (2023). Performance of ChatGPT on USMLE: Potential
jeehp.2023.20.5 for AI-assisted medical education using large lan­
Iskender, A. (2023). Holy or unholy? Interview with open guage models. PLoS Digital Health, 2(2), e0000198.
AI’s ChatGPT. European Journal of Tourism Research, https://doi.org/10.1371/journal.pdig.0000198

Page 15 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

Kuzman, T., Ljubešić, N., & Mozetič, I. (2023). ChatGPT: future perspectives and potential limitations.
Beginning of an end of manual annotation? Use case medRxiv: 2023.2002.2019.23286155.
of automatic genre identification. arXiv preprint Sanmarchi, F., Golinelli, D., & Bucci, A. (2023). A step-by-
arXiv:2303.03953. step Researcher’s Guide to the use of an AI-based
Lai, V. D., Ngo, N. T., Veyseh, A. P. B., Man, H., transformer in epidemiology: An exploratory analysis
Dernoncourt, F., Bui, T., & Nguyen, T. H. (2023). of ChatGPT using the STROBE checklist for observa­
Chatgpt beyond english: Towards a comprehensive tional studies. Journal of Public Health. medRxiv:
evaluation of large language models in multilingual 2023.2002.2006.23285514. https://doi.org/10.1007/
learning. arXiv preprint arXiv:2304.05613. s10389-023-01936-y
Lehnert, K. (2023). AI insights into theoretical physics Shahriar, S., & Hayawi, K. (2023). Let’s have a chat!
and the Swampland program: A journey through the A conversation with ChatGPT: Technology, applica­
cosmos with ChatGPT. arXiv preprint tions, and limitations. Artificial Intelligence and
arXiv:2301.08155. Applications. arXiv preprint arXiv:2302.13817. https://
Leiter, C., Zhang, R., Chen, Y., Belouadi, J., Larionov, D., doi.org/10.47852/bonviewAIA3202939
Fresen, V., & Eger, S. (2023). ChatGPT: A meta-analysis Sinha, R. K., Roy, A. D., Kumar, N., Mondal, H., & Sinha, R.
after 2.5 months. arXiv preprint arXiv:2302.13795. (2023). Applicability of ChatGPT in assisting to solve
Li, J., Dada, A., Kleesiek, J., & Egger, J. (2023). ChatGPT in higher order problems in pathology. Cureus, 15(2).
healthcare: A taxonomy and systematic review. https://doi.org/10.7759/cureus.35237
medRxiv: 2023.2003. 2030.23287899. Sobania, D., Briesch, M., Hanna, C., & Petke, J. (2023).
McGee, R. W. 2023. Who were the 10 best and 10 worst An analysis of the automatic bug fixing perfor­
US presidents? The Opinion of chat GPT (Artificial mance of ChatGPT. arXiv preprint
intelligence). The Opinion of Chat GPT (Artificial arXiv:2301.08653.
Intelligence). February 23, 2023. Sohail, S. S. (2023). A promising start and not a panacea:
Merow, C., Serra-Diaz, J. M., Enquist, B. J., & Wilson, A. M. ChatGPT’s early impact and potential in medical
(2023). AI chatbots can boost scientific coding. science and Biomedical Engineering research. Annals
Nature Ecology & Evolution. of Biomedical Engineering, 1–5. https://doi.org/10.
Murk, W., Goralnick, E., Brownstein, J. S., & Landman, A. B. 1007/s10439-023-03335-6
(2023). An opportunity to standardize and enhance Sohail, S. S., Farhat, F., Himeur, Y., Nadeem, M.,
Intelligent virtual assistant-delivered layperson cardi­ Madsen, D. Ø., Singh, Y., Atalla, S., & Mansoor, W.
opulmonary resuscitation instructions. medRxiv: (2023). Decoding ChatGPT: A taxonomy of existing
2023.2003.2009.23287050. research, Current challenges, and possible future
Orduña-Malea, E., & Cabezas-Clavijo, Á. (2023). ChatGPT directions. Journal of King Saud University - Computer
and the potential growing of ghost bibliographic and Information Sciences, 35(8), 101675. https://doi.
references. Scientometrics, 128(9), 5351–5355. org/10.1016/j.jksuci.2023.101675
https://doi.org/10.1007/s11192-023-04804-4 Srivastava, M. 2023. A day in the life of ChatGPT as
Pardos, Z. A., & Bhandari, S. (2023). Learning gain differ­ a researcher: Sustainable and efficient machine
ences between ChatGPT and human tutor generated learning-A review of sparsity techniques and future
algebra hints. arXiv preprint arXiv:2302.06871. research directions.
Paul, J., Ueno, A., & Dennis, C. (2023). ChatGPT and con­ Susnjak, T. (2022). ChatGPT: The end of online exam
sumers: Benefits, pitfalls and future research agenda. integrity?. arXiv preprint arXiv:2212.09292.
Wiley Online Library. Taecharungroj, V. (2023). What can ChatGPT do?” ana­
Pavlik, J. V. (2023). Collaborating with ChatGPT: lyzing early reactions to the innovative AI chatbot on
Considering the implications of Generative Artificial Twitter. Big Data and Cognitive Computing, 7(1), 35.
intelligence for Journalism and media education. https://doi.org/10.3390/bdcc7010035
Journalism & Mass Communication Educator, 78(1), Tlili, A., Shehata, B., Adarkwah, M. A., Bozkurt, A.,
84–93. https://doi.org/10.1177/10776958221149577 Hickey, D. T., Huang, R., & Agyemang, B. (2023). What
Rahimi, F., & Abadi, A. T. B. (2023). ChatGPT and pub­ if the devil is my guardian angel: ChatGPT as a case
lication ethics. Archives of Medical Research, 54(3), study of using chatbots in education. Smart Learning
272–274. https://doi.org/10.1016/j.arcmed.2023. Environments, 10(1), 15. https://doi.org/10.1186/
03.004 s40561-023-00237-x
Ray, P. P. (2023). ChatGPT: A comprehensive review on Ufuk, F. (2023). The role and limitations of large language
background, applications, key challenges, bias, models such as ChatGPT in clinical settings and
ethics, limitations and future scope. Internet of medical journalism. Radiology, 307(3), 230276.
Things and Cyber-Physical Systems, 3, 121–154. https://doi.org/10.1148/radiol.230276
https://doi.org/10.1016/j.iotcps.2023.04.003 Vaishya, R., Misra, A., & Vaish, A. (2023). ChatGPT: Is this
Rivas, P., & Zhao, L. (2023). Marketing with ChatGPT: version good for healthcare and research? Diabetes &
Navigating the ethical terrain of GPT-Based chatbot Metabolic Syndrome: Clinical Research & Reviews, 17
technology. AI, 4(2), 375–384. https://doi.org/10. (4), 102744. https://doi.org/10.1016/j.dsx.2023.
3390/ai4020019 102744
Rudolph, J., Tan, S., & Tan, S. (2023). ChatGPT: Bullshit Wang, J., Hu, X., Hou, W., Chen, H., Zheng, R., Wang, Y.,
spewer or the end of traditional assessments in higher Yang, L., Huang, H., Ye, W., & Geng, X. (2023). On the
education? Journal of Applied Learning & Teaching, 6 robustness of ChatGPT: An adversarial and out-of-
(1). https://doi.org/10.37074/jalt.2023.6.1.9 distribution perspective. arXiv preprint
Sajjad, M., & Saleem, R. (2023). Evolution of healthcare arXiv:2302.12095.
with ChatGPT: A word of caution. Annals of Wang, J., Liang, Y., Meng, F., Qu, Z., Li, J., & Zhou, J. (2023).
Biomedical Engineering, 51(8), 1663–1664. https:// Cross-Lingual Summarization via ChatGPT.
doi.org/10.1007/s10439-023-03225-x Transactions of the Association for Computational
Sallam, M. (2023). The utility of ChatGPT as an example of Linguistics 10, 1304–1323. arXiv preprint
large language models in healthcare education, arXiv:2302.14229. https://doi.org/10.1162/tacl_a_
research and practice: Systematic review on the 00520

Page 16 of 17
Gupta et al., Cogent Business & Management (2023), 10: 2275851
https://doi.org/10.1080/23311975.2023.2275851

Wang, A., McCarron, R., Azzam, D., Stehli, A., Xiong, G., & Accounting Education, 1–28. https://doi.org/10.2308/
DeMartini, J. (2022). Utilizing Big data from Google ISSUES-2023-013
trends to map population depression in the United Xue, V. W., Lei, P., & Cho, W. C. (2023). The potential
States: Exploratory infodemiology study. Journal of impact of ChatGPT in clinical and translational
Medical Internet Research Mental Health, 9(3), medicine. Clinical and Translational Medicine, 13(3).
e35253. https://doi.org/10.2196/35253 https://doi.org/10.1002/ctm2.1216
Wohlin, C. 2014. Guidelines for snowballing in sys­ Zarifhonarvar, A. (2023). Economics of ChatGPT: A labor
tematic literature studies and a replication in market view on the occupational impact of
software engineering. Paper presented at the Artificial intelligence. SSRN Electronic Journal.
Proceedings of the 18th international conference Available at SSRN 4350925. https://doi.org/10.2139/
on evaluation and assessment in software engi­ ssrn.4350925
neering (pp. 1–10). https://doi.org/10.1145/ Zhai, X. (2022). ChatGPT user experience: Implications for
2601248.2601268 education. SSRN Electronic Journal. Available at SSRN
Wong, I. A., Lian, Q. L., & Sun, D. (2023). Autonomous 4312418. https://doi.org/10.2139/ssrn.4312418
travel decision-making: An early glimpse into Zhang, B., Ding, D., & Jing, L. (2022). How would stance
ChatGPT and generative AI. Journal of Hospitality & detection techniques evolve after the launch of
Tourism Management, 56, 253–263. https://doi.org/ ChatGPT?. arXiv preprint arXiv:2212.14548.
10.1016/j.jhtm.2023.06.022 Zhu, Y., Han, D., Chen, S., Zeng, F., & Wang, C. 2023. How
Wood, D. A., Achhpilia, M. P., Adams, M. T., Aghazadeh, S., can ChatGPT benefit pharmacy: A case report on
Akinyele, K., Akpan, M., Allee, K. D., Allen, A. M., review writing.
Almer, E. D., & Ames, D. (2023). The ChatGpt Artificial Zhuo, T. Y., Huang, Y., Chen, C., & Xing, Z. (2023). Exploring
Intelligence Chatbot: How Well Does It Answer ai ethics of chatgpt: A diagnostic analysis. arXiv pre­
Accounting Assessment Questions? Issues in print arXiv:2301.12867.

Page 17 of 17