Generative AI: Exploring the future of innovative Automation and ethical Implications
ABSTRACT
Generative Artificial Intelligence (Gen AI) represents a transformative pressure inside the domain names of
creativity, design, content material era, and trouble-solving. This paper explores the advancements in Gen
AI models, along with the architectural improvements that electricity them and the societal implications they
bring. With packages spanning from artwork and literature to medical research and medical discovery,
the talents of Gen AI have exceeded conventional obstacles. but the fast increase of Gen
AI also raises ethical concerns about the capability misuse of AI-generated content,
deepfake generation, highbrow assets, and bias perpetuation. This paper delves into both the technical
underpinnings and the moral demanding situations, presenting a balanced angle on the promise and perils of
Gen AI.
1. INTRODUCTION
Generative artificial intelligence gear and packages (GenAI) structures robotically learn patterns and structures
from textual content, pictures, sounds, animation, fashions, or other media inputs to generate
new statistics with comparable traits [1]. GenAI is used to go looking, write, and create models, pc codes,
and artwork bureaucracy without human help. GenAI has emerged notably inside the contemporary decade
to help each industry thru exceptional products such as ChatGPT, Bing Chat, Bard, LLaMA, solid Diffusion,
Midjourney, and DALL-E [2-5]. nearly all industries proportion an constructive imaginative and prescient,
with extensive funding in the use of GenAI to transform aspects of cost chains [6-10]. however, similar
to many different technology hypes, whether or not this optimism will translate to value results or be a “fad
or style” stays to be examined through the years.
The adoption of GenAI in health care is emerging. studies factor to the usage of GenAI
in carrier interactions concerning breast most cancers diagnoses [11], bariatric surgery [12], cardiopulmonary
resuscitation [13], and breast cancer radiologic choice-making [14]. GenAI has the capability to convert by
using appearing responsibilities at better fine than people, which may also lessen errors associated
with people in expert domains including cancer detection [15] and neurological clinical selections [16].
The rise of GenAI is likewise called the “2nd system age” [17], whereby “instead
of machines acting mechanical work they're taking on cognitive paintings completely inside
the human domain” [17]. despite the fact that these instances are encouraging,
how precisely GenAI allows in fitness care tactics wishes to be articulated and evaluated
to offer an understanding of use and fee linkages [18,19].
2. Generative AI: Architectural Foundations
The wonderful abilties of Gen AI models are built upon numerous machine getting to
know architectures, each with its specific benefits:
2.1 Generative hostile Networks (GANs)
GANs, delivered through Ian Goodfellow in 2014, encompass neural networks: a generator and a
discriminator. The generator creates new data times, whilst the discriminator evaluates
them against actual facts. over time, the generator improves, developing data this is almost indistinguishable
from actual-international samples. GANs have
revolutionized picture era, main to advancements in creating practical photos, landscapes, and
even deepfake generation.
2.2 Transformer-primarily based models
Transformers, the architecture underlying models like GPT (Generative Pre-trained Transformer), have
dramatically changed the panorama of natural language generation. these models use self-
attention mechanisms to technique text, permitting them to apprehend context and generate coherent,
contextually applicable content. The emergence of GPT-four and past illustrates
how huge language fashions can generate essays, articles, code, and even creative writing
with minimum human input.
2.3 Variational Autoencoders (VAEs)
VAEs are another magnificence of generative models used mostly for image era and reconstruction. in contrast
to GANs, VAEs rent probabilistic strategies to generate statistics, focusing extra on encoding and
reconstructing enter statistics in new, unique ways. those fashions have observed packages in clinical imaging,
anomaly detection, and more.
3 Applications of Generative AI
Generative AI has spread across severa industries, offering progressive improvements and streamlining
workflows. some of the key applications consist of:
3.1 Innovative content technology
Generative AI has infiltrated innovative fields along with artwork, track, and writing. AI-
powered platforms like OpenAI’s DALL·E can generate images from textual descriptions, allowing designers
and artists to quickly visualize ideas. in the writing sphere, GPT-based fashions permit content introduction,
summarization, and language translation, offering new levels of productivity for content creators, reporters,
and researchers.
3.2 Healthcare and Drug Discovery
In healthcare, Gen AI is being employed to create artificial scientific facts, accelerate drug discovery,
and assist in prognosis. as an example, AI can generate molecular systems that align with
the homes wished for brand spanking new pills, significantly reducing the time required
for studies and improvement.
The effect of GenAI in fitness care depends on various factors, which
include the unique utility, great of information used for education, ethical issues, and regulatory framework
in vicinity. continuous monitoring, evaluation, and responsible deployment are critical to maximize the high-
quality effect and mitigate capability terrible effects. as an instance, synthetic intelligence (AI) assists
pathologists in diagnosing sicknesses from pathology slides, leading to quicker and more correct diagnoses
and improving affected person effects [20]. analysis of oncology literature, scientific trial records, and affected
person data can help oncologists pick out personalised, proof-based
totally remedy options for sufferers with cancer, potentially enhancing treatment selections [21]. AI has
been implemented to analyze medical images for situations including diabetic retinopathy, aiding in early
detection and intervention [22]. AI analyzes scientific and molecular facts to help physicians
make more knowledgeable choices about most cancers remedy and steer them in the direction
of personalized and effective treatment plans [23].
3.3 Sport layout and digital Environments
In gaming, Gen AI has been instrumental in automating the creation of characters, textures, and
even entire virtual worlds. AI-generated content enables builders to layout more immersive and
expansive virtual stories in much less time.
3.4 Information Augmentation and synthetic records
In gadget mastering, Gen AI can generate synthetic information to train fashions while real-
world information is scarce or luxurious to reap. This method is specially useful in regions consisting
of facial reputation, autonomous riding, and
cybersecurity, wherein categorized data is critical for constructing accurate models.
4 Moral Implications of Generative AI
Even as the capacity of Gen AI is big, it also brings with it numerous ethical and societal challenges.
4.1 Deepfakes and misinformation
one of the maximum contentious ethical issues around Gen AI is the creation of deepfakes, hyper-
practical artificial media that could mislead viewers. even as deepfakes can have benign makes use of, along
with in enjoyment and virtual artistry, additionally they pose serious risks for misinformation, fraud, and
political manipulation. Controlling the spread of AI-generated fake content has turn out to
be a international situation.
4.2 Intellectual assets and Copyright
As Gen AI can produce works that mimic existing data,
questions round possession and intellectual assets arise. Who owns the rights to AI-generated content
material? have to AI-generated artwork or literature be taken into
consideration original? these legal ambiguities venture present day copyright legal guidelines, making
it hard for creators, groups, and regulatory our bodies to adapt.
4.3 Bias and equity
AI fashions, inclusive of generative fashions, often inherit biases from the statistics they may be skilled on. If
the education information incorporates racial, gender, or socioeconomic biases, the AI can
perpetuate those biases in its outputs. This increases worries approximately equity in programs like
hiring, regulation enforcement, and even computerized content technology, in which biased
outputs should exacerbate inequality.
4.4 Impact on Human Creativity and labour
Gen AI’s capacity to automate content creation brings up worries about its impact on human creativity and
employment. while AI can help in innovative obligations, it additionally raises questions about the function of
human artists, writers, and designers in an increasingly computerized global. there may be a danger that
reliance on AI-generated content should diminish the cost of human
creativity, reducing possibilities in creative professions.
5 Future possibilities of Generative AI
Notwithstanding the ethical concerns, the future of Gen AI appears promising. because
the era matures, greater state-of-the-art fashions will emerge, capable
of appearing responsibilities with more nuance and
creativity. innovations in version education techniques, such as self-supervised getting to know and
reinforcement gaining knowledge of, will enable Gen AI to generate
even more complicated and significant outputs. moreover, collaborations between AI and human creators may
want to lead to hybrid kinds of creativity, wherein humans and machines work together to
produce unique works.
Efforts to address ethical concerns are also underway, with researchers and policymakers advocating for
clearer guidelines around AI utilization. those include frameworks for coping
with highbrow assets rights, strategies to discover and counter deepfakes,
and techniques to ensure AI fashions are educated on various, unbiased information.
Perceptron and Multilayer Perceptron (MLP)
Perceptron:
A Perceptron is the simplest type of neural network model, consisting of a single neuron. It works as a linear
classifier, where the input is multiplied by weights, summed, and passed through an activation function to
produce an output. The basic perceptron can only solve linearly separable problems.
Key Components of a Perceptron:
• Input Layer: Receives input features (x1, x2, ... xn).
• Weights (w1, w2, ... wn): Each input feature is associated with a weight.
• Bias: A constant added to the weighted sum to adjust the output.
• Activation Function: Determines the output by passing the weighted sum through a function, typically
a step function or sigmoid.
Multilayer Perceptron (MLP):
A Multilayer Perceptron (MLP) is a more complex version of a perceptron and is one of the simplest forms of
a feedforward neural network. It consists of an input layer, one or more hidden layers, and an output layer.
Each layer has multiple neurons, and neurons in one layer are fully connected to the next layer.
• Input Layer: Receives the input data.
• Hidden Layers: These layers allow MLP to learn complex patterns and relationships by applying
nonlinear transformations.
• Output Layer: Produces the final output, such as classification or regression.
Key Characteristics of MLP:
• Nonlinearity: MLP can solve non-linear problems due to the presence of hidden layers and non-linear
activation functions like ReLU or sigmoid.
• Universal Approximation: MLPs can approximate any continuous function with enough neurons and
hidden layers.
Understanding Bias and Variance
Bias and variance are key concepts in machine learning that refer to different sources of error in a model.
1. Bias:
• Definition: Bias refers to the error introduced by approximating a complex problem with a simplified
model. A model with high bias makes strong assumptions about the data and oversimplifies the problem.
• Impact:
o A high-bias model will likely underfit the data, meaning it performs poorly on both the training and test
datasets because it cannot capture the underlying patterns in the data.
o Example: A linear model trying to predict non-linear relationships will have high bias.
• Characteristics of High Bias:
o Simple models (e.g., linear regression for non-linear data).
o Underfitting the data.
o Low performance on both training and test data.
2. Variance:
• Definition: Variance refers to the model's sensitivity to small fluctuations in the training data. A model
with high variance captures noise and patterns specific to the training data but fails to generalize well to
unseen data.
• Impact:
o A high-variance model will likely overfit the data, performing very well on the training set but poorly
on the test set because it has memorized the training data rather than learning generalizable patterns.
o Example: A very deep neural network without regularization might overfit and have high variance.
• Characteristics of High Variance:
o Complex models (e.g., deep neural networks with too many parameters).
o Overfitting the data.
o Excellent performance on training data but poor performance on test data.
3. Bias-Variance Trade-off:
The bias-variance tradeoff is the balance between the two types of error:
• High Bias, Low Variance: The model is too simple and misses relevant relationships (underfitting).
• Low Bias, High Variance: The model is too complex and fits noise in the training data (overfitting).
The goal is to find a model with the optimal balance:
• Minimize total error by finding a tradeoff between bias and variance where the model performs well on
both training and test data.
Total Error:
Total error is the combination of bias, variance, and irreducible error:
Total Error=Bias2+Variance+Irreducible Error\text{Total Error} = \text{Bias}^2 + \text{Variance} + \
text{Irreducible Error}Total Error=Bias2+Variance+Irreducible Error
• Irreducible error: This is the error inherent in the problem that cannot be reduced by any model, such as
noise in the data.
Visualizing Bias-Variance Tradeoff
1. High Bias (Underfitting):
o the model is too simple, and predictions are far from the actual values.
o Both training and test errors are high because the model cannot capture the underlying structure.
2. High Variance (Overfitting):
o the model is too complex, fitting the training data perfectly but failing to generalize to new, unseen
data.
o Training error is very low, but test error is high due to overfitting.
3. Optimal Model:
o an optimal model strikes a balance between bias and variance, capturing the underlying pattern without
fitting the noise.
o Both training and test errors are relatively low, indicating good generalization.
Strategies to Manage Bias and Variance
1. To Reduce Bias (for Underfitting):
o Use more complex models (e.g., add more features, use deeper networks, or use non-linear models).
o Increase the number of hidden layers in neural networks.
o Decrease regularization (if applied too aggressively).
2. To Reduce Variance (for Overfitting):
o Use simpler models (e.g., reduce the number of features or neurons).
o Apply regularization techniques like L1 (Lasso), L2 (Ridge), or dropout in neural networks.
o Use cross-validation to select the best model and tune hyperparameters.
o Increase the amount of training data to help the model generalize better.
7. Legal and Regulatory Structures for Generative AI
As Generative AI advances in capability and sway, legal and regulatory structures have struggled to keep pace.
Governments and entities worldwide are striving to find equilibrium between fostering innovation and safegua
rding intellectual property (IP), privacy, and societal well-being.
7.1 Intellectual Property Rights (IPR)
One of the most urgent legal issues surrounding Generative AI is ownership. Who possesses a piece of content
crafted by an AI algorithm? Can an AI even be said to "own" its creations? Traditionally, copyright legislation
grants the creator authority over their intellectual property. However, in the realm of AI-generated works, the i
dentity of the creator becomes unclear. For example, if an AI model produces a novel or a work of art, should t
he developers who designed the AI, the users who input data, or the AI itself be granted the rights?
Different nations are approaching this situation in various ways. In the UK, the individual who sets up the AI t
o yield a specific outcome might hold ownership rights, while in the US, courts have largely ruled that AI cann
ot possess copyright. The establishment of universally recognized guidelines for IPR in AI-generated content i
s vital as this technology becomes more entwined with creative sectors.
7.2 Data Privacy and Protection
Generative AI models such as GPT and GANs are frequently trained on extensive datasets, which may encom
pass personal information. Ensuring compliance with privacy legislation, like the General Data Protection Reg
ulation (GDPR) in the European Union, is crucial. GDPR stipulates that individuals have rights concerning the
ir personal data, including the right to access, amend, or erase it. However, when an AI model utilizes this data
to forge synthetic content, it initiates questions regarding consent and data ownership.
An AI system trained on sensitive personal information may inadvertently produce content that reflects or exp
oses private details, posing considerable privacy threats. Regulatory bodies must guarantee that AI systems are
developed with stringent privacy protections, particularly when handling sensitive information.
7.3 Bias and Fairness Regulation
Numerous AI models inherit prejudices from their training data, potentially resulting in discriminatory outcom
es. Regulatory frameworks could be pivotal in mandating that organizations employing AI maintain fairness a
nd avoid discrimination in their AI systems. Developers need to adopt proactive measures, such as using diver
se datasets and implementing bias detection strategies, to prevent the continuation or intensification of societal
inequalities.
Given the societal ramifications of biased AI, enhanced regulatory oversight may be necessary to avert AI fro
m worsening gender, racial, and socio-economic disparities. Regions like the EU are rolling out regulations go
verning the ethical utilization of AI systems, and enhanced international collaboration may arise to standardize
these methods.
8. Ethical Considerations in Generative AI
The ethics of Generative AI are fundamentally connected to questions of responsibility, authority, and societal
influence. While this technology holds tremendous potential for positive impact, it also poses ethical dilemmas
8.1 AI's Role in Art and Creativity
Generative AI’s capability to produce artistic works has ignited debates about the essence of creativity. Human
creativity is frequently perceived as distinctive due to its subjective, emotional, and experiential attributes. Wh
en machines generate creative pieces, it challenges our comprehension of what it means to create.
For instance, AI-generated art has been auctioned for millions of dollars, prompting questions about the worth
of machine-created art as opposed to human-made. Furthermore, does AI truly "understand" creativity, or is it
merely generating patterns based on data?
8.2 Responsibility for AI Outputs
The inquiry of who is accountable for the actions of AI is vital, particularly when things go awry. For instance,
deepfake technology has been employed to fabricate fake videos and sway public perception, resulting in real-
world consequences. In these situations, it remains ambiguous whether the originators of the AI model, the use
rs, or the system itself should bear the blame.
Establishing clear delineations of responsibility is essential, especially for applications like autonomous weapo
nry or systems engaged in crucial decision
making in healthcare and justice. Ethical AI development should prioritize transparency, and users of such tech
nologies must be informed about the capabilities and constraints of AI systems.
9. The Prospects of Human-AI Cooperation
As Generative AI progresses, it is becoming evident that the future is rooted in partnership between humans an
d AI. Instead of AI superseding human creativity, the most effective applications are those where AI suppleme
nts and elevates human talents.
9.1 Blended Creativity
In disciplines such as art, design, and literature, AI has showcased its capability as a formidable instrument for
enhancing creativity. Creators are leveraging AI to produce innovative pieces of music, visual art, and narrativ
es that they may not have conceived independently. Rather than perceiving AI as an adversary, these artists reg
ard it as a collaborator—a companion that can help expand the scope of their creations.
The future may witness the emergence of "AI-enhanced creativity," where artists and AI systems collaborate to
forge new avenues of expression. This composite strategy could usher in the development of completely new a
rtistic styles and mediums, merging the realms of human and machine creativity.
9.2 Amplified Productivity
Beyond creativity, AI has proven its capacity to boost efficiency across various sectors. In areas such as conten
t formulation, journalism, and software engineering, Generative AI tools have the potential to automate tedious
tasks, enabling experts to concentrate on more intricate and meaningful endeavors. For example, AI systems c
an facilitate the creation of preliminary drafts of articles or reports, which humans can afterward refine and cus
tomize.
As AI becomes increasingly woven into everyday workflows, we may observe a rise in human-machine partne
rships, where AI manages the more routine responsibilities, allowing humans to apply their creativity, analytic
al skills, and critical thinking to higher-order problems.
10. Tackling the Hurdles of Scaling Generative AI
As Generative AI technologies expand, numerous challenges must be confronted to ensure sustainable and ethi
cal advancement. These obstacles encompass computational resource requirements, ecological repercussions, a
nd accessibility issues.
10.1 Computational Requirements
Training extensive Generative AI models calls for substantial computational resources. Models like GPT-4 und
ergo training on tens of thousands of GPUs for several months, consuming vast quantities of energy. This pro
mpts concerns regarding the environmental effects of AI technologies, as data centers already play a significan
t role in global energy usage.
Initiatives are in progress to design more energy-efficient AI models and algorithms that minimize the comput
ational footprint without compromising performance. Furthermore, organizations should contemplate shifting t
o renewable energy sources to power AI development, thereby diminishing their carbon footprint.
10.2 Broadening Access to AI
A fundamental promise of Generative AI is its ability to democratize content production. However, the resourc
es necessary to train and implement these models typically mean that only large corporations and well-
capitalized research institutions can afford to utilize them.
To avert the monopolization of AI technology by a select few, steps should be taken to ensure these tools are a
vailable to smaller enterprises, startups, and individuals. Open-source projects and cloud-based services that pr
ovide AI as a resource could play a crucial role in equalizing opportunities and ensuring that the advantages of
AI are broadly distributed.
. Conclusion
Generative AI represents a frontier of innovation so that it will keep to reshape a couple of industries.
Its packages, starting from creative content material era to scientific research, demonstrate its ability to
revolutionize how we consider introduction and automation. however, alongside its transformative talents, Gen
AI introduces ethical demanding situations that have to be addressed thoughtfully. As
we move toward a destiny an increasing number of shaped by means of generative models,
the balance among leveraging their ability and mitigating their dangers may be crucial
References
1. Pasick A. Artificial intelligence glossary: neural networks and other terms explained. The New York
Times. 2023. [2024-01-29]. [Link] .
2. Roose K. A coming-out party for generative A.I., Silicon Valley’s new craze. The New York
Times. 2022. Oct, [2024-01-29]. [Link] .
3. Karpathy A, Abeel P, Brockman G, Chen P, Cheung V, Duan Y. Generative models. Open AI. 2016.
[2024-01-31]. [Link] .
4. Metz C. OpenAI plans to up the ante in tech’s A.I. race. The New York Times. 2023. [2024-01-
29]. [Link]
%20the%20long%20term,Brockman%20said .
5. Thoppilan R, De Freitas D, Hall J, Shazeer N, Kulshreshtha A, Cheng HT, Jin A, Bis T, Baker L, Du Y,
Li Y, Lee H, Zheng HS, Ghafouri A, Menegali M, Huang Y. LaMDA: language models for dialog
applications. arXiv Preprint posted online January 20, 2022. 2020 doi: 10.48550/arXiv.2201.08239.
doi: 10.48550/arXiv.2201.08239. [CrossRef] [CrossRef] [Google Scholar]
6. Don’t fear an ai-induced jobs apocalypse just yet:
the west suffers from too little automation, not too much. The Economist. 2023. [2024-01-
29]. [Link]
.
7. Harreis H, Koullias T, Roberts R, Te K. Generative AI: unlocking the future of fashion. McKinsey &
Company. [2024-08-10]. [Link]
unlocking-the-future-of-fashion .
8. Eapen TT, Venkataswamy L, Finkenstadt DJ, Folk J. How generative AI can augment human
creativity. Harvard Business Review. 2023. [2024-01-29]. [Link]
can-augment-human-creativity .
9. The race of the AI labs heats up: ChatGPT is not the only game in town. The Economist. 2023. [2023-
01-30]. [Link] .
10. Google Cloud brings generative AI to developers, businesses, and governments. Google. 2023. [2024-
01-29]. [Link]
governments .
11. Zheng D, He X, Jing J. Overview of artificial intelligence in breast cancer medical imaging. J Clin
Med. 2023 Jan 04;12(2):419. doi: 10.3390/jcm12020419. [Link]
pii=jcm12020419 .jcm12020419 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
12. Samaan JS, Yeo YH, Rajeev N, Hawley L, Abel S, Ng WH, Srinivasan N, Park J, Burch M, Watson R,
Liran O, Samakar K. Assessing the accuracy of responses by the language model ChatGPT to questions
regarding bariatric surgery. Obes Surg. 2023 Jun 27;33(6):1790–6. doi: 10.1007/s11695-023-06603-
5. [Link] .10.1007/s11695-023-06603-5 [PMC free
article] [PubMed] [CrossRef] [Google Scholar]
13. Ahn C. Exploring ChatGPT for information of cardiopulmonary resuscitation. Resuscitation. 2023
Apr;185:109729. doi: 10.1016/[Link].2023.109729.S0300-9572(23)00042-4 [PubMed]
[CrossRef] [Google Scholar]
14. 14. Rao A, Kim J, Kamineni M, Pang M, Lie W, Succi MD. Evaluating ChatGPT as an adjunct for
radiologic decision-making. medRxiv. Preprint posted online February 7, 2023. 2023
doi: 10.1101/2023.02.02.23285399. doi: 10.1101/2023.02.02.23285399.2023.02.02.23285399
[CrossRef] [CrossRef] [Google Scholar]
15. 15. Cirillo D, Núñez-Carpintero I, Valencia A. Artificial intelligence in cancer research: learning at
different levels of data granularity. Mol Oncol. 2021 Apr;15(4):817–29. doi: 10.1002/1878-
0261.12920. [Link] . [PMC free article] [PubMed]
[CrossRef] [Google Scholar]
16. 16. Pedersen M, Verspoor K, Jenkinson M, Law M, Abbott DF, Jackson GD. Artificial intelligence for
clinical decision support in neurology. Brain Commun. 2020;2(2):fcaa096.
doi: 10.1093/braincomms/fcaa096. [Link] .fcaa096 [PMC free
article] [PubMed] [CrossRef] [Google Scholar]
17. 17. Brynjolfsson E, McAfee A. Second Machine Age: Work, Progress, and Prosperity in a Time of
Brilliant Technologies. New York, NY: WW Norton & Company; 2014. [Google Scholar]
18. Glum, J. (2015).Marketing to Generation Z: Millennials Move Aside as Brands Shift Focus to Under-
18 Customers. International Business News.
19. Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., &
Bengio, Y. (2014). Generative Adversarial Networks. arXiv preprint arXiv:1406.2661
20. Cascella M, Montomoli J, Bellini V, Bignami E. Evaluating the feasibility of ChatGPT in healthcare:
an analysis of multiple clinical and research scenarios. J Med Syst. 2023 Mar 04;47(1):33.
doi: 10.1007/s10916-023-01925-4. [Link] .10.1007/s10916-
023-01925-4 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
21. Baxi V, Edwards R, Montalto M, Saha S. Digital pathology and artificial intelligence in translational
medicine and clinical practice. Mod Pathol. 2022 Jan;35(1):23–32. doi: 10.1038/s41379-021-00919-
2. [Link] .S0893-3952(22)00347-7 [PMC
free article] [PubMed] [CrossRef] [Google Scholar]
22. Yu SH, Kim MS, Chung HS, Hwang EC, Jung SI, Kang TW, Kwon D. Early experience with Watson
for Oncology: a clinical decision-support system for prostate cancer treatment
recommendations. World J Urol. 2021 Feb;39(2):407–13. doi: 10.1007/s00345-020-03214-y.
doi: 10.1007/s00345-020-03214-y.10.1007/s00345-020-03214-y [PubMed] [CrossRef]
[CrossRef] [Google Scholar]
23. Wang Z, Keane PA, Chiang M, Cheung CY, Wong TY, Ting DS. Artificial intelligence and deep
learning in ophthalmology. In: Lidströmer N, Ashrafian H, editors. Artificial Intelligence in
Medicine. Cham, Switzerland: Springer; 2022. pp. 1519–52. [Google Scholar]
24. Jha, N., Shankar, P. R., Al-Betar, M. A., Mukhia, R., Hada, K., & Palaian, S. (2022). Undergraduate
medical students’ and interns’ knowledge and perception of artifcial intelligence in medicine. Advances
in Medical Education and Practice, 13, 927–937. [Link]
25. Krosnick, J. A., & Presser, S. (2015). Question and Questionnaire Design, Chapter 9. Handbook of
Survey Research. Kumar, V. V. R., & Raman, R. (2022). Student perceptions on Artifcial Intelligence
(AI) in higher education. In 2022 IEEE Integrated STEM Education Conference (ISEC).
[Link]
26. Lai, K. W., & Hong, K. S. (2015). Technology use and learning characteristics of students in higher
education: Do generational diferences exist? Br. J. Educ. Tech., 46(4), 725–738.
[Link]
27. Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-
native English writers. arXiv preprint arXiv:2304.02819.
28. Linnes, C., & Metcalf, B. R. (2017). iGeneration and their acceptance of technology. International
Journal of Management & Information Systems (IJMIS). [Link]