Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article presents a survey of cloud/edge computing that considers AI and its impact on IoT. It is well-written and has significant references in the field. My great concern is regarding novelty. The article does not clearly make a contribution. Other papers and books in the field have already extensively covered most concepts.

Following, I give some suggestions for improving the article:

- There is no methodology. How were the topics researched and considered in the article?

The introduction should clearly describe the scientific contribution regarding state-of-the-art.

- Figures are very common diagrams found in many books/articles on distributed systems.

The analysis of commercial solutions is interesting, but it is difficult to understand the differences between approaches. A more specific comparison of features instead of citing the specialized tools would be more meaningful. 

- Section 6 has interesting insights about the future that could be summarized in a scheme / table.

 

Author Response

We offer our sincere thanks to Reviewer 1 for the helpful comments. We have revised our paper according to the suggestions.

 

Comment 1: There is no methodology. How were the topics researched and considered in the article?

 

Response:  Section “Methodology” has been included discussing the research methodologies for the literature review in the paper. The section is as follows:

 

The purpose of this study was to look into relevant information regarding AI integration into cloud and edge computing, we focused on conducting a systematic literature review. We had an exhaustive search approach across various academic databases such as Google Scholar, IEEE Xplore, and ACM Digital Library. The targeted keywords were, in particular, "AI on Edge," "Edge Computing," "Cloud Computing," "AI for Edge," "Edge Intelligence," and "Cloud Service Providers." Using Boolean operators, attempts were then made to further develop the search by combining different search terms relevant to the subject matter, such as "AI AND Edge Computing" and "Cloud Computing OR Fog Computing." All these searches were restricted only to publications written in the English language so that research is focused on text written in one language.

 

We carefully selected the inclusion and exclusion criteria to ensure that only relevant and good-quality studies were included. To observe recent trends, we included only peer-reviewed articles in journals or conferences that appeared in major technical reports within the last five years. Studies relating to integrating AI in cloud and edge computing systems or proposing novel methodologies, frameworks, or that had contributed significantly to edge intelligence and AI-driven IoT applications were given a higher priority. Second, it included papers on cloud service providers and their AI-enabled services to understand the contribution from this area. We excluded non-peer-reviewed articles, white papers, unpublished theses, studies unrelated to the direct integration of AI in cloud or edge computing, and those articles older than five years unless seminal works in the field.

 

The review process underwent multiple stages to ensure that the identified papers underwent a rigorous assessment. First, the titles and abstracts of the identified papers were screened for relevance to our study. Those whose titles and abstracts did not correspond to our target inclusion criteria were excluded from further consideration. In the second stage, a complete text of each paper to be included based on the first screening was reviewed, which established a detailed assessment of methodologies applied, findings presented, and relevance to our research objectives. Finally, the quality and contribution of each paper to the field were evaluated. Those offering high-impact findings or introducing new approaches made it into the shortlist of works to be included. Extracted data for each study included AI integration methodologies in edge and cloud computing, findings on performance improvement, latency reduction, and energy efficiency, and applications that show practical implications of AI on Edge and AI for Edge.

 

Comment 2: The introduction should clearly describe the scientific contribution regarding state-of-the-art.

 

Response: The introduction section was extended with a discussion regarding previous works as follows:

 

Previous works conducted on the edge-cloud continuum focused on many aspects of the domain. Researchers in "Distributed Intelligence on the Edge-to-Cloud Continuum: A Systematic Literature Review" survey the current status of machine learning and data analytics frameworks, libraries, and paradigms enabling distributed intelligence across edge and cloud infrastructures. Challenges in ML workflow deployment upon such hybrid infrastructures are related to performance, reproducibility, and optimization of resources. The paper concludes by identifying open challenges in research and future directions toward the optimized deployment of AI workflows over heterogeneous edge-to-cloud environments. The paper "Disclosing Edge Intelligence: A Systematic Meta-Survey" surveys the Edge Intelligence paradigm proposed as an alternative solution to the limitations of cloud computing for services supporting IoT. It provides a systematic analysis of the literature available on EI concerning definitions, architectures, essential techniques, and future research directions. More specifically, the present study attempts to provide an overall picture for both experts and beginners, showing the present state, challenges, and possible future improvements of EI. Survey "Edge Intelligence—Research Opportunities for Distributed Computing Continuum Systems" covers opportunities and challenges about integrating edge computing with cloud computing in coming up with a DCCS. Then, it discusses how self-adaptive intelligence is required to manage the dynamic and heterogeneous nature of DCCS and how to use the MAPE-K framework. The paper identifies research opportunities and techniques that can help address the DCCS complexity and hopefully foster further development and collaboration.

 

Followed by the discussion we discussed our research focus on discussing the framework  “AI on Edge” and “AI for Edge” based on existing literature review and provided a concise summary of the structure and its different elements as follows:

 

In this paper, we give a fuller understanding of how AI not only enhances edge computing but also propels its evolution directly into a brand-new era of smart, independent systems capable of local decision-making. While edge computing and AI remain at the center of our concerns, attendant to it is the evolution of cloud service providers and their putative paths, particularly in terms of how their initiatives related to AI are remaking the horizon of the cloud. Based on our findings of the field we provide an image for the framework to implement the concept of AI for Edge and AI on Edge on different aspects of edge cloud. Furthermore, we entail a closer look at the market strategies and technological advancement such service providers have undergone in pursuit of their AI-powered services. It delivers an analytical insight into the changed scenario going on in cloud, fog, edge computing, and AI integration, whereby the authors focus on the significant role of IoT because it has sparked this revolution.

 

 

Comment 3: Section 6 has interesting insights about the future that could be summarized in a scheme / table.

Response: A new Table 3 discussing future of AI and clouds has been included. The table focuses on different aspects of AI in cloud, fog, edge, 5G, general trends, and regulatory and ethical concerns and what aspects of them the section discusses about. It also includes short detail on the research focus for each of the category.

 

 

 

In conclusion, your suggestions have been instrumental in refining our manuscript and enhancing its clarity, coherence, and overall presentation quality. We believe that these revisions address your concerns and strengthen the paper, making it a more valuable contribution to the field. We are thankful for your constructive feedback and the opportunity to improve our work.

Details are in the attached pdf for your perusal.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The topic is hot and the paper interesting but it is also flawed from several viewpoints.

My major concerns are about originality, methodology and presentation:

*no reference to computing continuum which is a key topic to spread intelligence from devices to cloud (see and cite<Pujol, Victor Casamayor, et al. "Edge intelligence—research opportunities for distributed computing continuum systems." IEEE Internet Computing 27.4 (2023): 53-74>, <Barbuto, Vincenzo, et al. "Disclosing edge intelligence: A systematic meta-survey." Big Data and Cognitive Computing 7.1 (2023): 44> and <Rosendo, Daniel, et al. "Distributed intelligence on the Edge-to-Cloud Continuum: A systematic literature review." Journal of Parallel and Distributed Computing 166 (2022): 71-94>). Explicitly discussing where smartness is located and the impact of this design choice is very relevant;

*Methodology is not clear: how Authors selected topics (e.g. which Research Questions?) and works (e.g. PRISMA approach)?

*As highlighted in <Barbuto, Vincenzo, et al. "Disclosing edge intelligence: A systematic meta-survey." Big Data and Cognitive Computing 7.1 (2023): 44> there are hundreds of surveys in this context: the added value of the proposal with respect to the existing literature should be further stressed;

*More graphical elements (such as comparison tables) should be inserted for enhance the effectiveness and readability and immediately display important hints;

*Conclusion lacks of effectiveness: take-aways and lessions learnt should be valorized, as well as ongoing and future research directions better highlighted;

Minor concerns:

*Missing paper outline at the end of Introduction;

*Proof-read the manuscript for fixing typos ("senor" or "infrastucture") and minor errors (e.g. "discuss about the"..Did you mean simply "discuss the"? Did you mean "comprises" or "consists of"?);

*Use past tense verbs for the conclusion;

*Some Figures are coarse-grained (e.g. Fig1)

*No reference to Generative ai: a comment might be inserted (see and cite <Morichetta, Andrea, Victor Casamayor Pujol, and Schahram Dustdar. "A roadmap on learning and reasoning for distributed computing continuum ecosystems." 2021 IEEE International Conference on Edge Computing (EDGE). IEEE, 2021.>)

Summarizing, the paper has some merit but in the current form cannot be accepted, in my opinion. A careful and deep revision phase is needed.

Comments on the Quality of English Language

Some edit required

Author Response

We offer our sincere thanks to Reviewer 2 for the helpful comments. We have revised our paper according to the suggestions.

 

Comment 1: No reference to computing continuum which is a key topic to spread intelligence from devices to cloud (see and cite<Pujol, Victor Casamayor, et al. "Edge intelligence—research opportunities for distributed computing continuum systems." IEEE Internet Computing 27.4 (2023): 53-74>, <Barbuto, Vincenzo, et al. "Disclosing edge intelligence: A systematic meta-survey." Big Data and Cognitive Computing 7.1 (2023): 44> and <Rosendo, Daniel, et al. "Distributed intelligence on the Edge-to-Cloud Continuum: A systematic literature review." Journal of Parallel and Distributed Computing 166 (2022): 71-94>). Explicitly discussing where smartness is located and the impact of this design choice is very relevant.

 

Response: Mentioned papers and their research focus has been discussed at the introduction following our research focus on contrary to these existing research works. The discussion is as follows:

 

 

Previous works conducted on the edge-cloud continuum focused on many aspects of the domain. Researchers in "Distributed Intelligence on the Edge-to-Cloud Continuum: A Systematic Literature Review" survey the current status of machine learning and data analytics frameworks, libraries, and paradigms enabling distributed intelligence across edge and cloud infrastructures. Challenges in ML workflow deployment upon such hybrid infrastructures are related to performance, reproducibility, and optimization of resources. The paper concludes by identifying open challenges in research and future directions toward the optimized deployment of AI workflows over heterogeneous edge-to-cloud environments. The paper "Disclosing Edge Intelligence: A Systematic Meta-Survey" surveys the Edge Intelligence paradigm proposed as an alternative solution to the limitations of cloud computing for services supporting IoT. It provides a systematic analysis of the literature available on EI concerning definitions, architectures, essential techniques, and future research directions. More specifically, the present study attempts to provide an overall picture for both experts and beginners, showing the present state, challenges, and possible future improvements of EI. Survey "Edge Intelligence—Research Opportunities for Distributed Computing Continuum Systems" covers opportunities and challenges about integrating edge computing with cloud computing in coming up with a DCCS. Then, it discusses how self-adaptive intelligence is required to manage the dynamic and heterogeneous nature of DCCS and how to use the MAPE-K framework. The paper identifies research opportunities and techniques that can help address the DCCS complexity and hopefully foster further development and collaboration.

In this paper, we give a fuller understanding of how AI not only enhances edge computing but also propels its evolution directly into a brand-new era of smart, independent systems capable of local decision-making. While edge computing and AI remain at the center of our concerns, attendant to it is the evolution of cloud service providers and their putative paths, particularly in terms of how their initiatives related to AI are remaking the horizon of the cloud. Based on our findings of the field we provide an image for the framework to implement the concept of AI for Edge and AI on Edge on different aspects of edge cloud. Furthermore, we entail a closer look at the market strategies and technological advancement such service providers have undergone in pursuit of their AI-powered services. It delivers an analytical insight into the changed scenario going on in cloud, fog, edge computing, and AI integration, whereby the authors focus on the significant role of IoT because it has sparked this revolution.

 

 

Comment 2: Methodology is not clear: how Authors selected topics (e.g. which Research Questions?) and works (e.g. PRISMA approach)?

 

Response: Methodology section has been included discussing the focus while formatting the paper and how the relevant papers were selected summarized as follows:

 

The purpose of this study was to look into relevant information regarding AI integration into cloud and edge computing, we focused on conducting a systematic literature review. We had an exhaustive search approach across various academic databases such as Google Scholar, IEEE Xplore, and ACM Digital Library. The targeted keywords were, in particular, "AI on Edge," "Edge Computing," "Cloud Computing," "AI for Edge," "Edge Intelligence," and "Cloud Service Providers." Using Boolean operators, attempts were then made to further develop the search by combining different search terms relevant to the subject matter, such as "AI AND Edge Computing" and "Cloud Computing OR Fog Computing." All these searches were restricted only to publications written in the English language so that research is focused on text written in one language.

 

We carefully selected the inclusion and exclusion criteria to ensure that only relevant and good-quality studies were included. To observe recent trends, we included only peer-reviewed articles in journals or conferences that appeared in major technical reports within the last five years. Studies relating to integrating AI in cloud and edge computing systems or proposing novel methodologies, frameworks, or that had contributed significantly to edge intelligence and AI-driven IoT applications were given a higher priority. Second, it included papers on cloud service providers and their AI-enabled services to understand the contribution from this area. We excluded non-peer-reviewed articles, white papers, unpublished theses, studies unrelated to the direct integration of AI in cloud or edge computing, and those articles older than five years unless seminal works in the field.

 

The review process underwent multiple stages to ensure that the identified papers underwent a rigorous assessment. First, the titles and abstracts of the identified papers were screened for relevance to our study. Those whose titles and abstracts did not correspond to our target inclusion criteria were excluded from further consideration. In the second stage, a complete text of each paper to be included based on the first screening was reviewed, which established a detailed assessment of methodologies applied, findings presented, and relevance to our research objectives. Finally, the quality and contribution of each paper to the field were evaluated. Those offering high-impact findings or introducing new approaches made it into the shortlist of works to be included. Extracted data for each study included AI integration methodologies in edge and cloud computing, findings on performance improvement, latency reduction, and energy efficiency, and applications that show practical implications of AI on Edge and AI for Edge.

 

Comment 3: More graphical elements (such as comparison tables) should be inserted for enhance the effectiveness and readability and immediately display important hints.

 

Response: Table 1 has been enhanced to provide more coherent summary of the discussion on the subject matter. Table 3 has been added to summarize the section “Paving The Future” highlighting the key points discussed in the section.

 

Comment 4: Conclusion lacks of effectiveness: take-aways and lessions learnt should be valorized, as well as ongoing and future research directions better highlighted.

 

 

Response: The conclusion section was rewritten focusing more on the highlighted points as follows:

 

In this paper, we have discussed how cloud computing, edge intelligence, and AI are converging to further transform the Internet-of-Things. The central insight from our study is that we have specified a framework for AI for edge and AI on edge by conducting a literature review. Computational power and scalability that complex AI and ML models require can be delivered through cloud computing, while hybrid cloud solutions enhance flexibility and deployment optimization. When linked with fog and edge computing, AI will supply real-time analytics for arriving data, better decision-making, and higher system efficiency. Edge AI reduces latency and enhances the efficiency and response time of IoT networks; this is rather crucial for those applications requiring real-time processing. Besides, 5G and MEC technologies support further real-time communication with reduced reliance on central servers.

Future research would hence wish to focus on developing standardized protocols and interoperability frameworks in such a rhythm that goes well with AI/ML applications across cloud, fog, and edge computing paradigms. It will be supplemented by next-generation AI hardware such as GPUs, TPUs, and NPUs for Edge AI performance. Second, edge-cloud orchestration techniques like federated learning and multi-agent systems will be essential; the assurance of ethical AI deployment is indispensable. Addressing these areas will fully harness the potential of cloud, edge, and AI convergence to create a brighter, more efficient, and secure digital future.

 

 

 

 

 

 

 

 

 

 

 

 

 

In conclusion, your suggestions have been instrumental in refining our manuscript and enhancing its clarity, coherence, and overall presentation quality. We believe that these revisions address your concerns and strengthen the paper, making it a more valuable contribution to the field. We are thankful for your constructive feedback and the opportunity to improve our work.

Details are in the attached pdf for your perusal.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript aims to provide a survey on the integration of artificial intelligence techniques within cloud and edge computing environments. It is well-written, and I have few recommendations for adjustments to the content. However, in my opinion, the manuscript lacks the necessary focus and depth required for publication in a scientific journal. 

A review article typically adopts one of two approaches: 1) a systematic review where the authors exhaustively search for all articles related to the topic under scrutiny and provide a logical systematization of the available literature, or 2) a survey where the authors discuss in a didactic/introductory tone the main aspects, algorithms, technologies, and architectures related to the topic under scrutiny. In my view, this manuscript achieves neither. There are some parts where the authors attempt to use the survey approach, but these are executed in a disjointed and very superficial manner. In my opinion, for the manuscript to be suitable for publication, it must either adopt a comprehensive and methodical approach to cover the literature exhaustively or provide a well-organized and thorough survey of the key topics within the scope.

Although Section 5 is interesting, I question its scientific relevance. The authors merely provide an overview of the functionalities of commercially available cloud services. This section occupies slightly more than one-third of the article. In my view, such a section could be included as an additional topic within the article but should not occupy a significant portion of the manuscript. 

An important aspect of a review or survey article is for the authors to identify existing research gaps and elaborate on the challenges that need to be addressed in the field. I consider that this paper does not satisfactorily present this aspect. Although there is a section titled "Paving The Future - Edge Computing and IoT," it primarily discusses already published articles and only superficially mentions future challenges. 

Upon finishing the article, I was left with the impression that I had only a superficial understanding of the topic addressed by the authors and that not all relevant aspects were mentioned. What I expect from a review article is the opposite: to finish reading with the sense that the topic has been thoroughly examined and that all pertinent and relevant aspects have been addressed and systematically presented. 

Author Response

We offer our sincere thanks to Reviewer 3 for the helpful comments. We have revised our paper according to the suggestions.

 

Comment 1: A review article typically adopts one of two approaches: 1) a systematic review where the authors exhaustively search for all articles related to the topic under scrutiny and provide a logical systematization of the available literature, or 2) a survey where the authors discuss in a didactic/introductory tone the main aspects, algorithms, technologies, and architectures related to the topic under scrutiny. In my view, this manuscript achieves neither. There are some parts where the authors attempt to use the survey approach, but these are executed in a disjointed and very superficial manner. In my opinion, for the manuscript to be suitable for publication, it must either adopt a comprehensive and methodical approach to cover the literature exhaustively or provide a well-organized and thorough survey of the key topics within the scope.

 

 

Response: Methodology section has been included discussing the research workflow as follows:

 

 

The purpose of this study was to look into relevant information regarding AI integration into cloud and edge computing, we focused on conducting a systematic literature review. We had an exhaustive search approach across various academic databases such as Google Scholar, IEEE Xplore, and ACM Digital Library. The targeted keywords were, in particular, "AI on Edge," "Edge Computing," "Cloud Computing," "AI for Edge," "Edge Intelligence," and "Cloud Service Providers." Using Boolean operators, attempts were then made to further develop the search by combining different search terms relevant to the subject matter, such as "AI AND Edge Computing" and "Cloud Computing OR Fog Computing." All these searches were restricted only to publications written in the English language so that research is focused on text written in one language.

 

We carefully selected the inclusion and exclusion criteria to ensure that only relevant and good-quality studies were included. To observe recent trends, we included only peer-reviewed articles in journals or conferences that appeared in major technical reports within the last five years. Studies relating to integrating AI in cloud and edge computing systems or proposing novel methodologies, frameworks, or that had contributed significantly to edge intelligence and AI-driven IoT applications were given a higher priority. Second, it included papers on cloud service providers and their AI-enabled services to understand the contribution from this area. We excluded non-peer-reviewed articles, white papers, unpublished theses, studies unrelated to the direct integration of AI in cloud or edge computing, and those articles older than five years unless seminal works in the field.

 

The review process underwent multiple stages to ensure that the identified papers underwent a rigorous assessment. First, the titles and abstracts of the identified papers were screened for relevance to our study. Those whose titles and abstracts did not correspond to our target inclusion criteria were excluded from further consideration. In the second stage, a complete text of each paper to be included based on the first screening was reviewed, which established a detailed assessment of methodologies applied, findings presented, and relevance to our research objectives. Finally, the quality and contribution of each paper to the field were evaluated. Those offering high-impact findings or introducing new approaches made it into the shortlist of works to be included. Extracted data for each study included AI integration methodologies in edge and cloud computing, findings on performance improvement, latency reduction, and energy efficiency, and applications that show practical implications of AI on Edge and AI for Edge.

 

 

Comment 2: Although Section 5 is interesting, I question its scientific relevance. The authors merely provide an overview of the functionalities of commercially available cloud services. This section occupies slightly more than one-third of the article. In my view, such a section could be included as an additional topic within the article but should not occupy a significant portion of the manuscript. 

 

Response:

The mentioned section of the paper is titled "Navigating the Commercial Cloud Ecosystem." It is devoted to providing a broad overview of the real-world cloud computing industry, mainly through the prism of fast-growing service providers: AWS, GCP, and Microsoft Azure. This section highlights practical implications and current trends in the commercial cloud landscape that set up the bigger context around cloud, edge, and AI technologies throughout the paper. This section presents an analysis of their services and innovations, offering a glimpse into how cloud computing is fast-tracking its entry into different industries. It also underscores how these providers are driving technological advancement, which makes the paper not only theoretically robust but also practically relevant. This contextual background enriches the understanding of the current market dynamics and technological innovations for the readers; its justification in a paper, therefore, makes sense. Moreover, Table 2 summarizing the discussion in the section has been updated for clearer understanding.

 

Comment 3: An important aspect of a review or survey article is for the authors to identify existing research gaps and elaborate on the challenges that need to be addressed in the field. I consider that this paper does not satisfactorily present this aspect. Although there is a section titled "Paving The Future - Edge Computing and IoT," it primarily discusses already published articles and only superficially mentions future challenges. 

Upon finishing the article, I was left with the impression that I had only a superficial understanding of the topic addressed by the authors and that not all relevant aspects were mentioned. What I expect from a review article is the opposite: to finish reading with the sense that the topic has been thoroughly examined and that all pertinent and relevant aspects have been addressed and systematically presented. 

 

Response: To better highlight the research work novelty we have discussed relevant works in the domain and also highlighted our research focus as opposed to existing literature as follows:

 

 

Previous works conducted on the edge-cloud continuum focused on many aspects of the domain. Researchers in "Distributed Intelligence on the Edge-to-Cloud Continuum: A Systematic Literature Review" survey the current status of machine learning and data analytics frameworks, libraries, and paradigms enabling distributed intelligence across edge and cloud infrastructures. Challenges in ML workflow deployment upon such hybrid infrastructures are related to performance, reproducibility, and optimization of resources. The paper concludes by identifying open challenges in research and future directions toward the optimized deployment of AI workflows over heterogeneous edge-to-cloud environments. The paper "Disclosing Edge Intelligence: A Systematic Meta-Survey" surveys the Edge Intelligence paradigm proposed as an alternative solution to the limitations of cloud computing for services supporting IoT. It provides a systematic analysis of the literature available on EI concerning definitions, architectures, essential techniques, and future research directions. More specifically, the present study attempts to provide an overall picture for both experts and beginners, showing the present state, challenges, and possible future improvements of EI. Survey "Edge Intelligence—Research Opportunities for Distributed Computing Continuum Systems" covers opportunities and challenges about integrating edge computing with cloud computing in coming up with a DCCS. Then, it discusses how self-adaptive intelligence is required to manage the dynamic and heterogeneous nature of DCCS and how to use the MAPE-K framework. The paper identifies research opportunities and techniques that can help address the DCCS complexity and hopefully foster further development and collaboration.

In this paper, we give a fuller understanding of how AI not only enhances edge computing but also propels its evolution directly into a brand-new era of smart, independent systems capable of local decision-making. While edge computing and AI remain at the center of our concerns, attendant to it is the evolution of cloud service providers and their putative paths, particularly in terms of how their initiatives related to AI are remaking the horizon of the cloud. Based on our findings of the field we provide an image for the framework to implement the concept of AI for Edge and AI on Edge on different aspects of edge cloud. Furthermore, we entail a closer look at the market strategies and technological advancement such service providers have undergone in pursuit of their AI-powered services. It delivers an analytical insight into the changed scenario going on in cloud, fog, edge computing, and AI integration, whereby the authors focus on the significant role of IoT because it has sparked this revolution.

 

 

The conclusion section has also been edited for be more focused on the research findings and summarize the points of our discussion and provide a future direction that this study will be helpful in providing in the discussion as follows:

 

In this paper, we have discussed how cloud computing, edge intelligence, and AI are converging to further transform the Internet-of-Things. The central insight from our study is that we have specified a framework for AI for edge and AI on edge by conducting a literature review. Computational power and scalability that complex AI and ML models require can be delivered through cloud computing, while hybrid cloud solutions enhance flexibility and deployment optimization. When linked with fog and edge computing, AI will supply real-time analytics for arriving data, better decision-making, and higher system efficiency. Edge AI reduces latency and enhances the efficiency and response time of IoT networks; this is rather crucial for those applications requiring real-time processing. Besides, 5G and MEC technologies support further real-time communication with reduced reliance on central servers.

Future research would hence wish to focus on developing standardized protocols and interoperability frameworks in such a rhythm that goes well with AI/ML applications across cloud, fog, and edge computing paradigms. It will be supplemented by next-generation AI hardware such as GPUs, TPUs, and NPUs for Edge AI performance. Second, edge-cloud orchestration techniques like federated learning and multi-agent systems will be essential; the assurance of ethical AI deployment is indispensable. Addressing these areas will fully harness the potential of cloud, edge, and AI convergence to create a brighter, more efficient, and secure digital future.

 

In conclusion, your suggestions have been instrumental in refining our manuscript and enhancing its clarity, coherence, and overall presentation quality. We believe that these revisions address your concerns and strengthen the paper, making it a more valuable contribution to the field. We are thankful for your constructive feedback and the opportunity to improve our work.

Details are in the attached pdf for your perusal.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

This paper provides a comprehensive overview of the convergence of AI, cloud computing, and edge intelligence. Overall, the authors carried on my suggestions for the first review. Revising it again, I still have some additional suggestions for improving the article:

 

1) The paper briefly mentions fog computing as an intermediary layer between the cloud and the edge. Given its growing importance in distributed computing, a more in-depth discussion of fog computing's role, architectures, and specific use cases would be valuable.

 

2)   While the paper touches upon security and privacy concerns, a more comprehensive discussion of the challenges and solutions in securing AI-powered cloud and edge systems would strengthen the paper. This could include topics like federated learning, differential privacy, and secure data transmission protocols.

 

3) The paper could benefit from a more critical discussion of the challenges and limitations of AI integration in cloud and edge computing. This could include issues like model complexity, resource constraints on edge devices, and the need for continuous learning and adaptation.

 

4) Some of the references seem dated. Including more recent research and publications would ensure the paper reflects the latest advancements in the field.

 

Author Response

We offer our sincere thanks to Reviewer 1 for the helpful comments. Please find responses in accordance with the comments provided by the reviewer:

 

Comment 1: The paper briefly mentions fog computing as an intermediary layer between the cloud and the edge. Given its growing importance in distributed computing, a more in-depth discussion of fog computing's role, architectures, and specific use cases would be valuable.

 

Response:  The review comment on putting more emphasis on fog computing is hereby highly appreciated. Our paper has essentially focused on the implementation of AI at Edge and AI for Edge, with an inclination towards discussing its recent developments, applications, and advantages when AI is integrated directly at the edge of the network. Although fog computing also plays a very important role in the distributed computing landscape, we focused more on edge computing and its interaction with AI due to its criticality to real-time data processing and subsequent decision-making in IoT applications.

Fog computing acts more as an intermediary layer but has its own set of architectures and use cases, which in fact are worthy of full coverage. It would have enlarged the scope of our paper beyond edge intelligence if an elaborate discussion on fog computing were to be included. What we intended to do was take our readers deep into those specific aspects of AI-driven edge computing, seen against its fast pace of evolution and rising relevance to applications requiring low latency and high efficiency.

 

Comment 2: While the paper touches upon security and privacy concerns, a more comprehensive discussion of the challenges and solutions in securing AI-powered cloud and edge systems would strengthen the paper. This could include topics like federated learning, differential privacy, and secure data transmission protocols.

 

 

Response:  In the paper, we try to enable the audience with a full overview of how AI integration into edge computing is going to revolutionize it in terms of applications, performance improvements, and practical implementations. The problems in security and privacy are touched upon in order to emphasize their importance; however, delving deep into these was outside the intended scope of our study.

We share the view that deeper investigations into the challenges and solutions concerning security and privacy will be relevant to federated learning, differential privacy, and secure protocols of data transmission. These themes are indeed very important in building further adoption and trust in AI-powered systems.

 

Federated Learning is mentioned several times in the context of its role in maintaining data privacy and avoiding data breaches by enabling decision-making based on local datasets without the need for central data transmission. The use of FL has been mentioned in Section 6, AI for Edge and Section 7, AI on Edge. We have also mentioned the necessity of studying differential privacy and secure transmission protocols as a possible future research scope.

 

Comment 3: The paper could benefit from a more critical discussion of the challenges and limitations of AI integration in cloud and edge computing. This could include issues like model complexity, resource constraints on edge devices, and the need for continuous learning and adaptation.

Response: While the primary focus of our paper was on the benefits and applications of AI in these paradigms, we did address several key challenges and limitations throughout the paper.

In Section 5.2, Edge Intelligence and Benefits, we discuss how AI enhances the efficiency of edge devices by reducing computational overhead, which is particularly important given the limited computing resources and energy constraints on these devices:

"The incorporation of AI into the cloud computing regime represents a recent paradigm shift that transforms the technology’s many core components. It combines the rapid evolution of computational intelligence and cloud infrastructure’s massive capacity to cut costs, automate resource management, boost systems reliability through predictive analytics, and guarantee new flexibility in privacy secured by new security sensor data insights from the system."

 "AI will make the edge more independent while also significantly enhancing it into a robust and versatile intelligent seat. Hence, the edge will evolve from an organization back into a system that is more user-centric and is able to provide customized solutions that dynamically match ever-changing user demands and needs."

"AI in edge has influenced different aspects of the service. The study [22] verifies that AI has pervaded the workings of edge computing in terms of performance, cost, privacy, efficiency, and reliability. It enhances the reduction of computational overhead on the resources when the low computation-intensive AI models are put to implementation at the edge."

In Section 6, AI for Edge, we explore the role of continuous learning and adaptation through techniques such as federated learning and reinforcement learning:

"Federated Learning (FL), basing its decision on the local dataset, can avoid data breaches and help in privacy maintenance because decisions can be made without recourse to any central server."

"Edge AI also utilizes reinforcement learning, on which its base is built, akin to how humans learn from the environment. The RL algorithms learn to make decisions by the interaction of actions with their feedback. Such a method becomes very handy in dynamic systems marked with changing states, for example, the traffic management system that requires the algorithm to learn real-time conditions."

In our conclusion, we identify the need for future research to address these challenges comprehensively, including standardized protocols, next-generation AI hardware, and orchestration techniques.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The Authors have mostly addressed my previous comments so I'm now happy with the revised version. It may be accepted, therefore, in my opinion. But please perform a careful proof-reading for fixing remaining minors

Comments on the Quality of English Language

Some small edits needed

Author Response

Response to Reviewer 2’s Comments

 

We offer our sincere thanks to Reviewer 2 for the helpful comments. Please find responses in accordance with the comments provided by the reviewer:

 

Comment 1: The Authors have mostly addressed my previous comments so I'm now happy with the revised version. It may be accepted, therefore, in my opinion. But please perform a careful proof-reading for fixing remaining minors.

 

 

Response: We appreciate the reviewer’s positive feedback and are glad to hear that our revisions have addressed your previous comments. We value your thorough review and are pleased that you find the revised version satisfactory for acceptance.

We will undertake a careful and detailed proof-reading process to fix any remaining minor errors and ensure the highest quality of our manuscript. This will include checking for typographical errors, grammatical issues, and ensuring consistency throughout the paper.

Thank you for your valuable feedback and for helping us improve our work. We look forward to the publication of our manuscript.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This submission is a resubmission of the previously reviewed manuscript. The authors have provided responses to my previous comments, and I appreciate their efforts in addressing my concerns. While the authors have made improvements to specific points of the manuscript, I find that the modifications do not adequately resolve the main issues I highlighted in my initial review. In my opinion, the manuscript still lacks the necessary focus and depth required for publication in a scientific journal.

Author Response

Response to Reviewer 3’s Comments

 

We offer our sincere thanks to Reviewer 3 for the helpful comments. Please find responses in accordance with the comments provided by the reviewer:

 

Comment 1: This submission is a resubmission of the previously reviewed manuscript. The authors have provided responses to my previous comments, and I appreciate their efforts in addressing my concerns. While the authors have made improvements to specific points of the manuscript, I find that the modifications do not adequately resolve the main issues I highlighted in my initial review. In my opinion, the manuscript still lacks the necessary focus and depth required for publication in a scientific journal.

 

Response: We appreciate the reviewer's continued engagement with our manuscript and the acknowledgment of the improvements we have made. We value your feedback and understand your concerns regarding the focus and depth of our manuscript. We would like to take this opportunity to clarify the scope and objectives of our study. Our paper primarily concerns the convergence of cloud computing, edge intelligence, and AI, with special emphasis on AI for Edge and on Edge. We sought to give a detailed framework of how AI can be fused into edge computing to enhance real-time data processing, decision-making, and general system efficiency. This focus, we believe, is necessary to address the needs of real-time IoT applications that begin assuming relevance in a fully technologically driven world like today.

 

Addressing the Key Points:

1. Security and Privacy:
• We have discussed Federated Learning (FL) as a significant method for enhancing privacy and security within AI-powered systems. FL allows for localized decision-making, thereby avoiding central data transmission and reducing privacy risks. We believe this discussion adequately covers an important aspect of security within the context of our study.
2. Fog Computing:
• While our primary focus is on edge computing, we have acknowledged the role of fog computing as an intermediary layer. Given our specific aim to delve into AI integration at the edge, we believe our current discussion strikes the right balance without diverging from the main objectives.
3. Challenges and Limitations:
• Our paper addresses challenges such as computational overhead, resource constraints, and continuous learning through the lens of AI for Edge and AI on Edge. We have highlighted how these challenges are mitigated by leveraging AI techniques, thus providing practical insights into overcoming these limitations.
4. Framework for AI on Edge and AI for Edge:
• We have provided a detailed framework that outlines the practical applications and benefits of integrating AI at the edge. This includes discussing various AI models, hardware requirements, and real-world use cases that demonstrate the effectiveness of our proposed framework.

 

 

In conclusion, your suggestions have been instrumental in refining our manuscript and enhancing its clarity, coherence, and overall presentation quality. We believe that these revisions address your concerns and strengthen the paper, making it a more valuable contribution to the field. We are thankful for your constructive feedback and the opportunity to improve our work.

Author Response File: Author Response.pdf