The AI Shadow War: SaaS vs. Edge Computing Architectures

Kevin J McNamara; Preeti Gupta; Rhea Pritham Marpu

REVIEW 5 major objections 5 minor 39 references

The AI Shadow War: SaaS vs. Edge Computing Architectures

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Edge AI claims a 10,000x energy advantage over cloud inference.

desk verdict A readable but advocacy-heavy position piece whose headline 10,000x efficiency claim is unsourced and internally contradicted by its own Table I. read the letter →

arxiv 2507.11545 v1 pith:VUYSLXOB submitted 2025-07-09 cs.DC cs.ETcs.NE

Rhea Pritham Marpu , Kevin J McNamara , Preeti Gupta This is my paper · ORCID

classification cs.DCcs.ETcs.NE

keywords edgeAISaaStest-timetrainingmixture-of-expertsenergyefficiencydataprivacyhybridedge-cloudarchitectureLandauer'sprinciple

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that decentralized edge AI—running models on phones, wearables, and IoT devices—is not merely a niche alternative to centralized SaaS AI but a fundamentally more efficient, private, and eventually dominant architecture. The headline evidence is an energy comparison: modern ARM processors and AI accelerators perform inference at about 100 microwatts, versus about 1 watt for equivalent cloud processing, a 10,000x gap. From that gap, plus latency, privacy, and cost advantages, the paper concludes that hybrid edge-cloud systems are inevitable. A sympathetic reader should care because the claim reframes AI's future around physics and economics rather than model quality alone.

What carries the argument

The paper's central mechanism is the energy-per-inference comparison between a cloud server GPU and an on-device NPU, quantified as 100 microwatts versus 1 watt (10,000x). It uses this ratio as the anchor for a broader cost, latency, privacy, and market analysis. On the model side, the argument leans on test-time training (adapting model parameters during inference) and mixture-of-experts (activating only a fraction of a large model's parameters per task, e.g., 21B of 236B in the DeepSeek-V2 model), with quantization and distillation making such models feasible on edge hardware. The paper also invokes Landauer's principle—the minimum energy $kT\ln 2$ per erased bit—as a theoretical floor that, it argues, favors reducing data transmission and redundant computation.

What would settle it

Run one representative inference task (for example, a small language model or image classifier) and measure total system energy on an edge device—device power draw including memory and radios—versus the energy attributed to one cloud inference including data-center overhead and network transmission; if the full-system edge number is not orders of magnitude below the cloud number, the 10,000x claim does not hold.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that edge AI's efficiency advantage is so large that it changes the architecture argument. It claims modern ARM processors and specialized AI accelerators consume roughly 100 microwatts for inference while equivalent cloud processing consumes about 1 watt, a 10,000x difference; that local processing delivers 5–10 ms latency versus 100–500 ms for cloud round-trips; and that distributed processing removes single points of failure. It further claims that recent model innovations—test-time training and mixture-of-experts—let small on-device models close the performance gap with large cloud models. The paper then reads Landauer's principle as physical confirmation that minimizing data movement and redundant bit operations aligns edge computing with thermodynamic limits. The conclusion is that the conflict ends in hybrid ecosystems, not victory for either side.

Load-bearing premise

The argument depends on treating a physics minimum for erasing one bit as proof that spreading computation across devices uses less energy, even though the paper never connects that per-bit limit to actual data-transmission costs or total bit operations.

Editorial extensions

If this is right

If the 10,000x efficiency gap holds, energy costs for AI inference shift dramatically toward on-device processing, making battery-powered AI a default rather than an exception.
Privacy architecture changes: personal health and biometric data can be processed locally, so large-scale breach events like the 2023 healthcare breach cited in the paper become less likely by design.
Model design priorities change: test-time training, mixture-of-experts, and distillation become first-class techniques because they determine what can run on a $100–$200 device.
Market forecasts follow: the paper projects the edge AI market growing from $9B in 2025 to $49.6B in 2030, with hybrid architectures covering over 60% of major-sector deployments by 2035.
Policy would need energy-efficiency standards and data-sovereignty rules, as the paper proposes, to keep pace with the shift.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial inference: The 10,000x figure compares an on-device NPU to a server GPU doing equivalent inference, but a real end-to-end comparison would need to include the cloud's amortized data-center overhead and the edge device's full-system power; the paper does not make that accounting explicit.
Editorial inference: If Landauer's principle is the grounding, the decisive quantity is total bit erasures across the whole system, and the paper does not derive a placement-dependent energy bound; one testable extension would be a formal model comparing transmission energy to local compute energy.
Editorial inference: The hybrid conclusion could be tested by measuring whether real deployments shift toward on-device inference as NPUs reach a given TOPS-per-watt threshold, or whether cloud GPUs keep advancing faster.
Editorial inference: The market projections are presented as a consequence of the efficiency and privacy advantages, but they are exogenous figures in the paper; connecting the two would require a model that translates energy savings into adoption curves.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

What you should know first: this is a position essay, not a research paper. It surveys known edge-AI trends (TTT, MoE, distillation, NPUs) and assembles them into an argument that hybrid edge-cloud is inevitable. The qualitative points about privacy, latency, and offline capability are sound and reasonably well-illustrated with real incidents and market figures. The paper is clearly written and honest in places—Table I is explicitly labeled 'illustrative,' and Section V admits that a 236B model won't run on a tablet.

The soft spots are real and load-bearing. The central claim—a 10,000x efficiency advantage from 100 microwatts versus 1 watt for cloud inference—is unsourced and contradicted by Table I, which lists 1–10 mW for on-device NPUs (a 100x–1000x gap, not 10,000x). The abstract's '100 microwatts' is ten to a hundred times lower than the table's own range. No citation to a real measurement appears; the surrounding references [8], [23], [35] are about federated learning, autonomous driving, and sustainable processing, none of which contain this specific figure. The Landauer-principle argument in Section V.A is rhetorical: it states a physical minimum per bit erasure but never connects that limit to where computation physically sits, so the 'inevitability' conclusion rests on an interpretive leap, not a derivation. The TTT complexity O(n log n) is asserted without proof. The policy numbers (1W per billion ops by 2027, $500M funds) are invented specifics that no evidence grounds. Reference [36] points to OpenAI pricing as a source for inference cost per 1M tokens, which is at least checkable, but the comparison is crude.

On proportionality: the paper's qualitative architecture argument survives even if you throw out the 10,000x figure and the Landauer framing—edge AI does offer meaningful efficiency, latency, and privacy benefits. But the manuscript presents that argument with numbers that don't hold up, and the conclusion of inevitability leans on those numbers. That makes it a weak research submission, though a decent magazine article or industry blog post.

Who gets value: someone wanting a quick overview of the edge-vs-cloud talking points and a collection of relevant citations. Not someone needing reliable quantitative evidence.

My recommendation: desk reject for a research venue, but suggest the authors resubmit if they replace the efficiency claims with sourced, reproducible measurements and dial back the thermodynamic inevitability framing. I would not send it to a rigorous referee in its current form.

Referee Report

5 major / 5 minor

Summary. The paper is a position/advocacy piece arguing that edge AI is on track to outperform cloud SaaS AI in energy efficiency, privacy, and latency, and that hybrid edge-cloud architectures are inevitable. Its quantitative backbone is a purported 10,000x efficiency advantage (100 microwatts vs 1 watt per inference), a Landauer-principle argument for thermodynamic inevitability, and a series of market and regulatory projections. The paper applies these claims to education, healthcare, smart homes, autonomous transport, and speculative 2035 scenarios.

Significance. If the central efficiency and thermodynamic claims were substantiated, the paper would provide a strong quantitative case for accelerating edge AI investment and hybrid architecture policy. The paper usefully catalogs qualitative trade-offs and correctly identifies that hybrid architectures are a plausible near-term evolution. However, the quantitative claims are not supported by the manuscript's own data or by the cited literature; there are no machine-checked proofs, reproducible code, or parameter-free derivations. The stated contribution therefore remains an opinionated synthesis rather than a research result.

major comments (5)

[Abstract and §II.C.2; Table I] The central efficiency claim is internally inconsistent and unsupported. The Abstract and §II.C.2 state that modern ARM processors perform inference at 100 microwatts versus 1 watt for equivalent cloud processing, a 10,000x advantage. Table I, however, gives Edge AI energy per inference as ~1-10 mW and Big SaaS AI as ~1-10 W, implying a factor of 100x-10,000x; Section III explicitly says the table is 'illustrative' and shows 'the scale of difference rather than absolute benchmarks.' No cited source reports the 100 microwatts figure; the surrounding citations [8], [23], [35] concern federated learning, autonomous driving, and sustainable AI processing without quoting this measurement. Because Section VIII.A calls the 'documented 10,000x efficiency gains' an 'existential necessity,' this number is load-bearing and must be reconciled with the table or removed.
[§V.A] The Landauer-principle argument is a non-sequitur. After stating that at 300 K the minimum energy per bit erasure is about 2.9 × 10^-21 J, the paper concludes that edge AI's distributed approach 'fundamentally aligns with the physics of efficient information processing' and 'inherently reduce[s] the total number of bit operations required across the AI ecosystem.' No quantitative link is given between the per-bit Landauer bound and the placement of computation; data transmission and local computation involve different bit-operation counts, and moving computation to the edge could increase total operations if edge devices are less compute-efficient per operation. The claimed reduction in total bit operations is asserted, not derived, so the thermodynamic inevitability conclusion in §VIII.A is unsupported.
[§V.A] The complexity claim underlying the feasibility of test-time training is unproven. The paper states 'TTT operations scale as O(n log n) for parameter updates' without defining n (parameters, tokens, hidden dimension?) or providing a citation or derivation. This O(n log n) claim is used to conclude that TTT is 'feasible on edge devices with specialized hardware accelerators,' so it is load-bearing for the performance-equality claim; as written it is an unsupported assertion.
[§VI.B and §VII.A] The market and regulatory projections are presented without sources or methodology, and at least one is internally inconsistent. The paper projects growth from $9B (2025) to $49.6B (2030) at a 38.5% CAGR, but 9 × (1.385)^5 ≈ 45.9B, not 49.6B. The regulatory limits in §VII.A ('below 1W per billion operations by 2027, implemented through graduated limits: 5W (2025), 2W (2026), 1W (2027)') mix units and lack any basis in cost or feasibility analysis. These figures feed the 'Economic Imperative' and policy recommendations in §VIII.A, so they require either explicit sources or a stated modeling assumption.
[Introduction and §V.B; refs. [13], [28]] Several performance claims are misreferenced. The Introduction credits 'DeepSeek-Coder-V2 achieving high accuracy (79.8%) on AIME [28]', but reference [28] is the DeepSeek-R1 model card, which reports a different model; Section V.B describes 'DeepSeek-V2' and cites [13], which is the Switch Transformers paper, not a DeepSeek-V2 technical report. The model names also differ between sections (DeepSeek-Coder-V2 vs DeepSeek-V2). This makes the paper's opening evidence for edge AI performance unverifiable from the cited sources.

minor comments (5)

[Abstract] The phrase '100 microwatts forinference' is missing a space; it should be '100 microwatts for inference'.
[Section III, Table I] The 'Inference Cost' row lists an improvement factor of '>1,000x' for a comparison between ~$5-15 and <$0.01; at the lower bound this is 500x, so the stated factor should be justified or stated as a range.
[Section VIII.B] In the 'Ian' scenario, the on-premise rack consumes less than 500 watts and is called a 'tenfold reduction' relative to a cloud GPU rack consuming 'hundreds of kilowatts'; those numbers imply a 200-600x reduction, not 10x.
[Section V.A] The text says 'At room temperature (300K), this theoretical minimum is approximately 2.9 × 10^-21 joules per bit operation.' This is correct only if 'bit operation' means 'bit erased'; consider rephrasing to avoid confusion.
[Table I] Table I lists sources as '[8], [23], [26], [35]–[37]', but [36] (OpenAI pricing) and [37] (EIA electricity prices) do not directly support the per-inference energy figures; please add direct references or explain the inference.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found; the paper's unsupported 10,000x efficiency claim and rhetorical Landauer argument are correctness concerns, not circularity.

full rationale

The paper does not fit parameters to data or derive a quantitative prediction from its own inputs. The central efficiency figure ('10,000x efficiency advantage: modern ARM processors consume merely 100 microwatts for inference versus 1 watt for equivalent cloud processing', Abstract; repeated in Section II.C.2) is asserted without a supporting measurement and is internally inconsistent with Table I, which lists 'Energy per Inference' as '∼1-10W (Server GPU)' versus '∼1-10mW (On-Device NPU)', implying a range closer to 100–10,000x. The table itself says 'The values demonstrate the scale of difference rather than absolute benchmarks,' so the claim is illustrative and unsupported rather than derived. The Landauer's-principle discussion (Sections V.A and VIII.A) is rhetorical: it assumes that edge AI's distributed approach means 'minimizing unnecessary data movement and redundant computations' and then invokes Landauer's principle as 'confirm[ing]' that this 'fundamentally aligns with the physics of efficient information processing.' No equation connects the per-bit erasure energy kT ln 2 to the placement of computation, so the conclusion is an interpretive leap, not a circular derivation. There are no load-bearing self-citations (the authors' prior work is not invoked), no imported uniqueness theorem, and no ansatz smuggled in via citation. Accordingly, no step in the paper is equivalent by construction to a fitted input or to a self-referential definition; the appropriate finding is no significant circularity.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The paper has no mathematical derivation, so there are no conventionally fitted parameters. The numbers listed above are asserted constants that carry the argument's weight, either without sources or with sources that do not fully support them. The axioms capture the interpretive leap about Landauer's principle and the willingness to treat illustrative figures as evidence.

free parameters (5)

Edge inference power figure = 100 µW (abstract), 1-10 mW (Table I)
The core 10,000x efficiency claim rests on an asserted power number that is unsourced and internally inconsistent with the paper's own table.
Cloud inference power figure = 1 W (abstract), 1-10 W (Table I)
Baseline for the efficiency ratio, asserted without a benchmark definition or citation.
TTT computational complexity = O(n log n)
Stated in Section V.A without derivation or definition of n; used to argue feasibility of on-device test-time training.
Edge AI market CAGR = 38.5% (2025-2030)
Cited to a market report [30]; the implied CAGR from $9B to $49.6B over 5 years is approximately 40.7%, so the stated rate is internally inconsistent.
Regulatory power limits = 5W/2W/1W per billion operations by 2027
Proposed in Section VII.A with no feasibility, cost, or enforcement analysis; these are invented constants supporting the policy framework.

assumptions (3)

ad hoc to paper Landauer's principle supports the conclusion that distributed edge processing is thermodynamically advantageous
Section V.A invokes kT ln 2 as confirming edge AI's alignment with physics, but the principle only bounds bit-erasure energy and no model connects it to data center versus device placement.
domain assumption The cited performance and market datasets are accurate and representative
The paper's conclusions rest on external market reports and unsourced performance numbers, including the 100 µW inference claim and the India wearable trial.
domain assumption Illustrative numbers in Table I are meaningful for comparison
The table caption admits values 'demonstrate the scale of difference rather than absolute benchmarks', yet the paper uses these values as evidence for claims like 10,000x.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The AI Shadow War: SaaS vs. Edge Computing Architectures." pith.science (2026). https://pith.science/paper/VUYSLXOB

@misc{pith2026250711545,
  author       = {Pith},
  title        = {Pith review of: The AI Shadow War: SaaS vs. Edge Computing Architectures},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VUYSLXOB}},
  note         = {Machine review of arXiv:2507.11545}
}

abstract

The very DNA of AI architecture presents conflicting paths: centralized cloud-based models (Software-as-a-Service) versus decentralized edge AI (local processing on consumer devices). This paper analyzes the competitive battleground across computational capability, energy efficiency, and data privacy. Recent breakthroughs show edge AI challenging cloud systems on performance, leveraging innovations like test-time training and mixture-of-experts architectures. Crucially, edge AI boasts a 10,000x efficiency advantage: modern ARM processors consume merely 100 microwatts forinference versus 1 watt for equivalent cloud processing. Beyond efficiency, edge AI secures data sovereignty by keeping processing local, dismantling single points of failure in centralized architectures. This democratizes access throughaffordable hardware, enables offline functionality, and reduces environmental impact by eliminating data transmission costs. The edge AI market projects explosive growth from $9 billion in 2025 to $49.6 billion by 2030 (38.5% CAGR), fueled by privacy demands and real-time analytics. Critical applications including personalized education, healthcare monitoring, autonomous transport, and smart infrastructure rely on edge AI's ultra-low latency (5-10ms versus 100-500ms for cloud). The convergence of architectural innovation with fundamental physics confirms edge AI's distributed approach aligns with efficient information processing, signaling the inevitable emergence of hybrid edge-cloud ecosystems.

Figures

Figures reproduced from arXiv: 2507.11545 by the authors.

**Figure 2.** Distributed Edge(local processing on individual devices, no internet [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** User Perception of Data Security in SaaS vs. Edge AI Models. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 34 canonical work pages

[8]

Federated machine learning: Concept and applications,

Q. Yang, Y . Liu, T. Chen, and Y . Tong, “Federated machine learning: Concept and applications,” ACM Trans. Intell. Syst. Technol. , vol. 10, no. 2, Feb. 2019. [Online]. Available: https://dl.acm.org/doi/10.1145/ 3298981

work page 2019
[23]

Edge Computing for Real- Time Decision Making in Autonomous Driving: Review of Challenges, Solutions, and Future Trends

Xie, J.; Zhou, X.; Cheng, L. “Edge Computing for Real- Time Decision Making in Autonomous Driving: Review of Challenges, Solutions, and Future Trends.” Int. J. Adv. Comput. Sci. Appl. , vol. 15, no. 7, 2024, pp. 598–604. [Online]. Available: https://thesai.org/Downloads/V olume15No7/Paper 59-Edge Computing for Real Time Decision Making.pdf

work page 2024
[35]

Sustainable AI Processing at the Edge,

S. Ollivier et al., “Sustainable AI Processing at the Edge,” IEEE Micro, vol. 43, no. 1, pp. 19–28, Jan.-Feb. 2023. [Online]. Available: https: //ieeexplore.ieee.org/abstract/document/9941196

work page arXiv 2023
[36]

Pricing,

OpenAI, “Pricing,” OpenAI, n.d. [Online]. Available: https://openai.com/ pricing

work page
[28]

DeepSeek-R1,

DeepSeek-AI, “DeepSeek-R1,” Hugging Face, May 2025. [Online]. Available: https://huggingface.co/deepseek-ai/DeepSeek-R1

work page 2025
[13]

Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity,

W. Fedus, B. Zoph, and N. Shazeer, “Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity,” Journal of Machine Learning Research , vol. 23, no. 120, pp. 1–39, 2022. [Online]. Available: https://arxiv.org/abs/2101.03961

arXiv 2022
[1]

A survey on transfer learning,

S. J. Pan and Q. Yang, “A survey on transfer learning,” IEEE Trans. Knowl. Data Eng., vol. 22, no. 10, pp. 1345–1359, Oct. 2010. [Online]. Available: https://ieeexplore.ieee.org/document/5288526

arXiv 2010
[2]

Goodfellow, Y

I. Goodfellow, Y . Bengio, and A. Courville, Deep Learning. MIT Press,

work page

Show all 39 references

[3]

In-datacenter performance analysis of a tensor processing unit,

N. P. Jouppi et al., “In-datacenter performance analysis of a tensor processing unit,” in Proc. 44th Annu. Int. Symp. Comput. Archit. , 2017. [Online]. Available: https://dl.acm.org/doi/10.1145/3079856.3080246

2017
[4]

Membership inference attacks against machine learning models,

R. Shokri et al., “Membership inference attacks against machine learning models,” in IEEE S&P, 2017. [Online]. Available: https://ieeexplore.ieee. org/document/7958568

2017
[5]

Attention is all you need,

A. Vaswani et al., “Attention is all you need,” in NeurIPS,
[6]

Quantization and training of neural networks for effi- cient integer-arithmetic-only inference,

B. Jacob et al., “Quantization and training of neural networks for effi- cient integer-arithmetic-only inference,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018, pp. 2704–2713. [Online]. Available: https://ieeexplore.ieee...

2018
[7]

Energy and policy consider- ations for deep learning in NLP,

E. Strubell, A. Ganesh, and A. McCallum, “Energy and policy consider- ations for deep learning in NLP,” in Proc. 57th Annual. Meeting Assoc. Computer Linguistics , 2019. [Online]. Available: https://aclanthology. org/P19-1355/

2019
[9]

Russell and P

S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach , 4th ed. Pearson, 2020. [Online]. Available: https://aima.cs.berkeley.edu/

2020
[10]

40,000+ Internet-connected Cameras Exposed Streaming Live on The Internet,

“40,000+ Internet-connected Cameras Exposed Streaming Live on The Internet,” Cybersecurity News , 2025. [Online]. Available: https: //cybersecuritynews.com/40000-internet-connected-cameras-exposed/

2025
[11]

Bringing artificial intelligence to business management,

S. Feuerriegel, Y . R. Shrestha, G. von Krogh, and C. Zhang, “Bringing artificial intelligence to business management,” Nature Machine Intelli- gence, vol. 4, no. 8, pp. 611–613, Jul. 2022

2022
[12]

Regulatory frameworks for connected and autonomous vehicles: A systematic literature review and future research agenda,

N. Mouter, G. H. de Almeida Correia, and C. G. Chorus, “Regulatory frameworks for connected and autonomous vehicles: A systematic literature review and future research agenda,” Transportation Research Part A: Policy and Practice , vol. 165, pp. 234–251, Nov. 2022

2022
[14]

The energy consumption of large language models: A review,

J. M. S. Albalawi et al., “The energy consumption of large language models: A review,” in IEEE Conf. Smart Grid Green Energy , 2023. [Online]. Available: https://ieeexplore.ieee.org/document/10234567

2023
[15]

Data Breach Locations, Types, and Associated Characteristics Among US Hospitals,

M. Hufstader Gabriel, A. Noblin, A. Rutherford, A. Walden, and K. Cortelyou-Ward, “Data Breach Locations, Types, and Associated Characteristics Among US Hospitals,” Am. J. Manag. Care , vol. 24, no. 2, pp. 78–84, 2018

2018
[16]

December 2023 Healthcare Data Breach Report,

HIPAA Journal, “December 2023 Healthcare Data Breach Report,” 2023. [Online]. Available: https://www.hipaajournal.com/ december-2023-healthcare-data-breach-report/

2023
[17]

AI and the Digital Divide: Challenges and Opportunities

Lazanyuk, I.; Eyeberdiyeva, M.; Diaz, M. “AI and the Digital Divide: Challenges and Opportunities.” Springer, New York, 2025, pp. 283–288

2025
[18]

Adaptive Learning Systems: A Comprehensive Overview and Identification of Challenges

Hilali, K.; Chergui, M.; Ammoumou, A. “Adaptive Learning Systems: A Comprehensive Overview and Identification of Challenges.” in 2023 IEEE International Conference on Technology Management, Operations and Decisions (ICTMOD) , Rabat, Morocco, 2023, pp. 192–197

2023
[19]

Advancing Transformative Education: Gener- ative AI as a Catalyst for Equity and Innovation

Bura, C.; Myakala, P. K. “Advancing Transformative Education: Gener- ative AI as a Catalyst for Equity and Innovation.” arXiv, 2024. [Online]. Available: https://arxiv.org/abs/2411.15971

2024 arXiv
[20]

Wearable AI to enhance pa- tient safety and clinical decision-making

Mahajan, A.; Heydari, K.; Powell, D. “Wearable AI to enhance pa- tient safety and clinical decision-making.” npj Digit. Med. , vol. 8, no. 1, 176, 2025. [Online]. Available: https://www.nature.com/articles/ s41746-025-01554-w

2025
[21]

Edge AI Infer- ence in Heterogeneous Constrained Computing: Feasibility and Oppor- tunities

Morabito, R.; Tatipamula, M.; Tarkoma, S.; Chiang, M. “Edge AI Infer- ence in Heterogeneous Constrained Computing: Feasibility and Oppor- tunities.” in 2023 IEEE 28th International Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CAMAD), pp. ...

2023 arXiv
[22]

Edge Computing and its Application in Robotics: A Survey

Tahir, N.; Parasuraman, R. “Edge Computing and its Application in Robotics: A Survey.” arXiv, 2025. [Online]. Available: https://arxiv.org/ abs/2507.00523

2025 arXiv
[24]

Trust in AI: progress, challenges, and future directions

Afroogh, S.; Akbari, A.; Malone, E.; Kargar, M.; Alambeigi, H. “Trust in AI: progress, challenges, and future directions.” Humanities and Social Sciences Communications , vol. 11, no. 1, 1568, 2024. [Online]. Available: https://www.nature.com/articles/s41599-024-04044-8

2024
[25]

Standards for edge AI interoperability,

IEEE Standards Association, “Standards for edge AI interoperability,” IEEE Std 2857-2023, 2023. [Online]. Available: https://standards.ieee. org/ieee/2857/10634/

2023
[26]

HCA Healthcare Data Breach Impacts 11.27 Million Individuals,

HIPAA Journal, “HCA Healthcare Data Breach Impacts 11.27 Million Individuals,” HIPAA Journal, Jul. 2023. [Online]. Available: https: //www.hipaajournal.com/hca-healthcare-cyberattack-data-breach-2023/

2023
[27]

How to Stop Smart-Camera Hack- ers,

Consumer Reports, “How to Stop Smart-Camera Hack- ers,” January 2024. [Online]. Available: https://www. consumerreports.org/home-garden/home-security-cameras/ keep-home-security-cameras-from-being-hacked-a2927068390/

2024
[29]

Open-source AI models for edge deployment,

Hugging Face, “Open-source AI models for edge deployment,” 2025. [Online]. Available: https://huggingface.co/docs

2025
[30]

2024 State of Edge AI Report,

DATEurope, “2024 State of Edge AI Report,” May 2023. [Online]. Available: https://dateurope.com/wp-content/uploads/2024/05/ 2024STAGEOFEDGEAIREPORT.pdf

2024
[31]

Realizing the Full Potential of Edge AI with Con- nected Security,

Arm Newsroom, “Realizing the Full Potential of Edge AI with Con- nected Security,” Aug. 15, 2024. [Online]. Available: https://newsroom. arm.com/blog/psa-certified-2024-security-report

2024
[32]

2024 SaaS Disruption Report: Security and Data,

Onymos and ESG, “2024 SaaS Disruption Report: Security and Data,” Aug. 21, 2024. [Online]. Available: https://onymos.com/blog/ 2024-saas-disruption-report-security-and-data/

2024
[33]

The State of SaaS Security Report 2025- 2026,

Cloud Security Alliance, “The State of SaaS Security Report 2025- 2026,” Apr. 21, 2025. [Online]. Available: https://cloudsecurityalliance. org/artifacts/state-of-saas-security-report-2025

2025
[34]

2024 Connected Consumer Survey: Increasing Consumer Privacy and Security Concerns,

Deloitte, “2024 Connected Consumer Survey: Increasing Consumer Privacy and Security Concerns,” Jan. 2024. [Online]. Available: https://www.deloitte.com/us/en/about/press-room/ increasing-consumer-privacy-and-security-concerns-in-the-generative-ai-era. html

2024
[37]

Electricity explained: Factors affecting electricity prices,

U.S. Energy Information Administration (EIA), “Electricity explained: Factors affecting electricity prices,” U.S. Energy Information Ad- ministration, Jun. 29, 2023. [Online]. Available: https://www.eia.gov/ energyexplained/electricity/prices-and-factors-affecting-prices.php

2023
[2016]

Available: https://www.deeplearningbook.org/

[Online]. Available: https://www.deeplearningbook.org/
[2017]

Available: https://papers.nips.cc/paper/ 7181-attention-is-all-you-need.pdf

[Online]. Available: https://papers.nips.cc/paper/ 7181-attention-is-all-you-need.pdf

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [8]

Federated machine learning: Concept and applications,

Q. Yang, Y . Liu, T. Chen, and Y . Tong, “Federated machine learning: Concept and applications,” ACM Trans. Intell. Syst. Technol. , vol. 10, no. 2, Feb. 2019. [Online]. Available: https://dl.acm.org/doi/10.1145/ 3298981

work page 2019

[2] [23]

Edge Computing for Real- Time Decision Making in Autonomous Driving: Review of Challenges, Solutions, and Future Trends

Xie, J.; Zhou, X.; Cheng, L. “Edge Computing for Real- Time Decision Making in Autonomous Driving: Review of Challenges, Solutions, and Future Trends.” Int. J. Adv. Comput. Sci. Appl. , vol. 15, no. 7, 2024, pp. 598–604. [Online]. Available: https://thesai.org/Downloads/V olume15No7/Paper 59-Edge Computing for Real Time Decision Making.pdf

work page 2024

[3] [35]

Sustainable AI Processing at the Edge,

S. Ollivier et al., “Sustainable AI Processing at the Edge,” IEEE Micro, vol. 43, no. 1, pp. 19–28, Jan.-Feb. 2023. [Online]. Available: https: //ieeexplore.ieee.org/abstract/document/9941196

work page arXiv 2023

[4] [36]

Pricing,

OpenAI, “Pricing,” OpenAI, n.d. [Online]. Available: https://openai.com/ pricing

work page

[5] [28]

DeepSeek-R1,

DeepSeek-AI, “DeepSeek-R1,” Hugging Face, May 2025. [Online]. Available: https://huggingface.co/deepseek-ai/DeepSeek-R1

work page 2025

[6] [13]

Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity,

W. Fedus, B. Zoph, and N. Shazeer, “Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity,” Journal of Machine Learning Research , vol. 23, no. 120, pp. 1–39, 2022. [Online]. Available: https://arxiv.org/abs/2101.03961

arXiv 2022

[7] [1]

A survey on transfer learning,

S. J. Pan and Q. Yang, “A survey on transfer learning,” IEEE Trans. Knowl. Data Eng., vol. 22, no. 10, pp. 1345–1359, Oct. 2010. [Online]. Available: https://ieeexplore.ieee.org/document/5288526

arXiv 2010

[8] [2]

Goodfellow, Y

I. Goodfellow, Y . Bengio, and A. Courville, Deep Learning. MIT Press,

work page