Application Domains Research Articles

Supplemental Material for Fears About Artificial Intelligence Across 20 Countries and Six Domains of Application

Supplemental Material for Fears About Artificial Intelligence Across 20 Countries and Six Domains of Application

An Integrative, Systematic Review of the Situational Judgment Test Literature

Situational judgment tests (SJTs) are popular assessment approaches that present scenarios describing situations that one may experience in a job. Due to its long history and cross-disciplinary nature, today’s SJT literature is quite fragmented. In this integrative review, we start by systematically taking stock and synthesizing the SJT literature from the different scientific disciplines via bibliometric techniques on 524 unique documents. We identify six literature clusters (i.e., SJTs in the medical sciences, SJTs in personnel selection, methodological issues and SJTs for specific constructs, SJTs to assess emotional intelligence and related constructs, technological advances in SJTs, SJTs for teacher assessment and development) that correspond to academic disciplines and research streams within them. We also identify current trends in SJT research by examining the clusters formed by a recent subset of the SJT literature. We then build on the bibliometric analysis by categorizing the identified themes in an organizing framework with two fundamental dimensions: the main purpose of a study (i.e., conceptual understanding, prediction, other [e.g., understanding mean group differences, applicant reactions]) and its research focus (i.e., SJTs holistically, content, and design and methods). Finally, on the basis of this framework, we provide recommendations to encourage greater knowledge sharing between scientific disciplines. In addition, we outline an agenda for future research in terms of four broad directions: SJT theory, SJT constructs, SJT design and methods, and SJT application domains.

Characterizing the Performance and Cost of Blockchains on the Cloud and at the Edge

While state-of-the-art permissioned blockchains can achieve thousands of transactions per second on commodity hardware with x86/64 architecture, their performance when running on different architectures, such as ARM, is not clear. The goal of this work is to characterize the performance and cost of permissioned blockchains on different hardware systems, which is important as diverse application domains are adopting blockchains. To this end, we conduct extensive cost and performance evaluation of two permissioned blockchains, namely Hyperledger Fabric and ConsenSys Quorum, on five different types of hardware covering both x86/64 and ARM architecture, as well as both cloud and edge computing. The hardware nodes include servers with Intel Xeon CPU, servers with ARM-based Amazon Graviton CPU, and edge devices with ARM-based CPU such as Nvidia Jetson TX2 and Raspberry Pi 4. Our results reveal a diverse profile of the two blockchains across different settings, demonstrating the impact of hardware choices on the overall performance and cost. We find that Graviton servers outperform Xeon servers in many settings due to their powerful CPU and high memory bandwidth. Edge devices with ARM architecture, on the other hand, exhibit low performance. When comparing the cloud with the edge, we show that the cost of the latter is much smaller in the long run if manpower cost is not considered.

LesionScanNet: dual-path convolutional neural network for acute appendicitis diagnosis.

Acute appendicitis is an abrupt inflammation of the appendix, which causes symptoms such as abdominal pain, vomiting, and fever. Computed tomography (CT) is a useful tool in accurate diagnosis of acute appendicitis; however, it causes challenges due to factors such as the anatomical structure of the colon and localization of the appendix in CT images. In this paper, a novel Convolutional Neural Network model, namely, LesionScanNet for the computer-aided detection of acute appendicitis has been proposed. For this purpose, a dataset of 2400 CT scan images was collected by the Department of General Surgery at Kanuni Sultan Süleyman Research and Training Hospital, Istanbul, Turkey. LesionScanNet is a lightweight model with 765K parameters and includes multiple DualKernel blocks, where each block contains a convolution, expansion, separable convolution layers, and skip connections. The DualKernel blocks work with two paths of input image processing, one of which uses 3 × 3 filters, and the other path encompasses 1 × 1 filters. It has been demonstrated that the LesionScanNet model has an accuracy score of 99% on the test set, a value that is greater than the performance of the benchmark deep learning models. In addition, the generalization ability of the LesionScanNet model has been demonstrated on a chest X-ray image dataset for pneumonia and COVID-19 detection. In conclusion, LesionScanNet is a lightweight and robust network achieving superior performance with smaller number of parameters and its usage can be extended to other medical application domains.

Rational Engineering of a Dynamic, Enzyme-Driven DNA Walker for Intracellular Dual-Enzyme Activity Sequentially Monitoring and Imaging.

Monitoring enzyme activity is crucial in both scientific research and clinical applications. However, abnormalities in a single enzyme's activity can indicate multiple diseases, limiting the specificity of single enzyme activity monitoring in clinical diagnosis. We developed a dynamic DNA walker that can be sequentially activated by two enzymes, enabling the monitoring and imaging of both enzyme activities within cells. Initially, the DNA walker contains a site for apurinic/apyrimidinic endonuclease 1 (APE1). Upon APE1 activation, the DNA walker forms specific structures recognized and cleaved by Flap endonuclease 1 (FEN1). The temporal disparity between the activities of APE1 and FEN1 allows for the sequential monitoring and imaging of both enzymes, reducing the likelihood of false-positive results. To enhance local concentration and decrease reaction time, the DNA walk sequence was attached to the surface of gold nanoparticles (AuNPs). The fruition of this endeavor will facilitate the investigation and advancement of multiple enzyme activity monitoring and imaging methods and technologies, while simultaneously broadening the domains of application for DNA nanotechnology.

Sensors, Techniques, and Future Trends of Human-Engagement-Enabled Applications: A Review

Human engagement is a vital test research area actively explored in cognitive science and user experience studies. The rise of big data and digital technologies brings new opportunities into this field, especially in autonomous systems and smart applications. This article reviews the latest sensors, current advances of estimation methods, and existing domains of application to guide researchers and practitioners to deploy engagement estimators in various use cases from driver drowsiness detection to human–robot interaction (HRI). Over one hundred references were selected, examined, and contrasted in this review. Specifically, this review focuses on accuracy and practicality of use in different scenarios regarding each sensor modality, as well as current opportunities that greater automatic human engagement estimation could unlock. It is highlighted that multimodal sensor fusion and data-driven methods have shown significant promise in enhancing the accuracy and reliability of engagement estimation. Upon compiling the existing literature, this article addresses future research directions, including the need for developing more efficient algorithms for real-time processing, generalization of data-driven approaches, creating adaptive and responsive systems that better cater to individual needs, and promoting user acceptance.

The Transformative Patent Landscape in Saudi Arabia Since the Saudi Vision 2030 Announcement

This study analyzes the patent landscape of Saudi Arabia from the announcement of Saudi Vision 2030 in late April 2016 to September 2024, utilizing the Patsnap database to evaluate patent grants across various organizations. The findings reveal a gradual increase in patent registrations, with Saudi Aramco leading in patent grants, followed by King Faisal University (KFU), King Fahd University of Petroleum and Minerals (KFUPM), King Abdullah University of Science and Technology (KAUST), and King Abdulaziz University (KAU). SABIC, a prominent industry player in Saudi Arabia, has registered most of its patents using its European Head Office address and holds extensive collaborations with international partners, generating numerous patents. The analysis identifies the top patent offices where KSA organizations seek protection, including the United States Patent and Trademark Office (USPTO), Saudi Authority for Intellectual Property (SAIP), European Patent Office (EPO), the China National Intellectual Property Administration (CNIPA), and the German Patent and Trade Mark Office. However, the limited number of registrations at the SAIP highlights a need for improvement. The primary application domains encompass borehole/well accessories, measurement devices, organic chemistry, computing, and chemical/physical processes. The landscape reveals that Saudi Aramco and KFUPM focus predominantly on upstream and downstream technologies, while KAU, KFU, and KAUST concentrate on life sciences. Key findings indicate a significant increase in patent activity since the vision announcement, suggesting a growing focus on innovation within Saudi Arabia. However, the concentration of patents among a few major players (Saudi Aramco and SABIC) and the underrepresentation of patents filed with the Saudi Authority of Intellectual Property (SAIP) highlight areas for improvement. This study emphasizes the necessity to support small and medium enterprises (SMEs) and healthcare research institutions to foster broader participation in innovation and protect novel technologies. This research contributes valuable insights into the current state of patenting activities in Saudi Arabia and outlines opportunities for enhancing the country’s innovation ecosystem.

A Latency Composition Analysis for Telerobotic Performance Insights Across Various Network Scenarios

Telerobotics involves the operation of robots from a distance, often using advanced communication technologies combining wireless and wired technologies and a variety of protocols. This application domain is crucial because it allows humans to interact with and control robotic systems safely and from a distance, often performing activities in hazardous or inaccessible environments. Thus, by enabling remote operations, telerobotics not only enhances safety but also expands the possibilities for medical and industrial applications. In some use cases, telerobotics bridges the gap between human skill and robotic precision, making the completion of complex tasks requiring high accuracy possible without being physically present. With the growing availability of high-speed networks around the world, especially with the advent of 5G cellular technologies, applications of telerobotics can now span a gamut of scenarios ranging from remote control in the same room to robotic control across the globe. However, there are a variety of factors that can impact the control precision of the robotic platform and user experience of the teleoperator. One such critical factor is latency, especially across large geographical areas or complex network topologies. Consequently, military telerobotics and remote operations, for example, rely on dedicated communications infrastructure for such tasks. However, this creates a barrier to entry for many other applications and domains, as the cost of dedicated infrastructure would be prohibitive. In this paper, we examine the network latency of robotic control over shared network resources in a variety of network settings, such as a local network, access-controlled networks through Wi-Fi and cellular, and a remote transatlantic connection between Finland and the United States. The aim of this study is to quantify and evaluate the constituent latency components that comprise the control feedback loop of this telerobotics experience—of a camera feed for an operator to observe the telerobotic platform’s environment in one direction and the control communications from the operator to the robot in the reverse direction. The results show stable average round-trip latency of 6.6 ms for local network connection, 58.4 ms when connecting over Wi-Fi, 115.4 ms when connecting through cellular, and 240.7 ms when connecting from Finland to the United States over a VPN access-controlled network. 255,235,61. These findings provide a better understanding of the capabilities and performance limitations of conducting telerobotics activities over commodity networks, and lay the foundation of our future work to use these insights for optimizing the overall user experience and the responsiveness of this control loop.

Eeg based smart emotion recognition using meta heuristic optimization and hybrid deep learning techniques

In the domain of passive brain-computer interface applications, the identification of emotions is both essential and formidable. Significant research has recently been undertaken on emotion identification with electroencephalogram (EEG) data. The aim of this project is to develop a system that can analyse an individual’s EEG and differentiate among positive, neutral, and negative emotional states. The suggested methodology use Independent Component Analysis (ICA) to remove artefacts from Electromyogram (EMG) and Electrooculogram (EOG) in EEG channel recordings. Filtering techniques are employed to improve the quality of EEG data by segmenting it into alpha, beta, gamma, and theta frequency bands. Feature extraction is performed with a hybrid meta-heuristic optimisation technique, such as ABC-GWO. The Hybrid Artificial Bee Colony and Grey Wolf Optimiser are employed to extract optimised features from the selected dataset. Finally, comprehensive evaluations are conducted utilising DEAP and SEED, two publically accessible datasets. The CNN model attains an accuracy of approximately 97% on the SEED dataset and 98% on the DEAP dataset. The hybrid CNN-ABC-GWO model achieves an accuracy of approximately 99% on both datasets, with ABC-GWO employed for hyperparameter tuning and classification. The proposed model demonstrates an accuracy of around 99% on the SEED dataset and 100% on the DEAP dataset. The experimental findings are contrasted utilising a singular technique, a widely employed hybrid learning method, or the cutting-edge method; the proposed method enhances recognition performance.

Comparing human-labeled and AI-labeled speech datasets for TTS

As the output quality of neural networks in the fields of automatic speech recognition (ASR) and text-to-speech (TTS) continues to improve, new opportunities are becoming available to train models in a weakly supervised fashion, thus minimizing the manual effort required to annotate new audio data for supervised training. While weak supervision has recently shown very promising results in the domain of ASR, speech synthesis has not yet been thoroughly investigated regarding this technique despite requiring the equivalent training dataset structure of aligned audio-transcript pairs. In this work, we compare the performance of TTS models trained using a well-curated and manually labeled training dataset to others trained on the same audio data with text labels generated using both grapheme- and phoneme-based ASR models. Phoneme-based approaches seem especially promising, since even for wrongly predicted phonemes, the resulting word is more likely to sound similar to the originally spoken word than for grapheme-based predictions. For evaluation and ranking, we generate synthesized audio outputs from all previously trained models using input texts sourced from a selection of speech recognition datasets covering a wide range of application domains. These synthesized outputs are subsequently fed into multiple state-of-the-art ASR models with their output text predictions being compared to the initial TTS model input texts. This comparison enables an objective assessment of the intelligibility of the audio outputs from all TTS models, by utilizing metrics like word error rate and character error rate. Our results not only show that models trained on data generated with weak supervision achieve comparable quality to models trained on manually labeled datasets, but can outperform the latter, even for small, well-curated speech datasets. These findings suggest that the future creation of labeled datasets for supervised training of TTS models may not require any manual annotation but can be fully automated.

Machine learning materials properties with accurate predictions, uncertainty estimates, domain guidance, and persistent online accessibility

Abstract One compelling vision of the future of materials discovery and design involves the use of machine learning (ML) models to predict materials properties and then rapidly find materials tailored for specific applications. However, realizing this vision requires both providing detailed uncertainty quantification (model prediction errors and domain of applicability) and making models readily usable. At present, it is common practice in the community to assess ML model performance only in terms of prediction accuracy (e.g. mean absolute error), while neglecting detailed uncertainty quantification and robust model accessibility and usability. Here, we demonstrate a practical method for realizing both uncertainty and accessibility features with a large set of models. We develop random forest ML models for 33 materials properties spanning an array of data sources (computational and experimental) and property types (electrical, mechanical, thermodynamic, etc). All models have calibrated ensemble error bars to quantify prediction uncertainty and domain of applicability guidance enabled by kernel-density-estimate-based feature distance measures. All data and models are publicly hosted on the Garden-AI infrastructure, which provides an easy-to-use, persistent interface for model dissemination that permits models to be invoked with only a few lines of Python code. We demonstrate the power of this approach by using our models to conduct a fully ML-based materials discovery exercise to search for new stable, highly active perovskite oxide catalyst materials.

Multi-Degree-of-Freedom Platforms: Development and Outlook

Background: Multi-degree-of-freedom platforms are frequently employed in space docking devices, motion modeling, and robotics. Complex research is being done on multidegree- of-freedom platforms, including work on spatial route creation, control simulation, and forward and backward platform motion solutions. The development of research platforms in their current state is advantageous for advancing multi-degree-of-freedom motion research, which has the potential to advance numerous sectors through new technology. Objective: In this patent study, the benefits and drawbacks of various platforms are explored to analyze the development trend of platforms for academics and engineers through the most recent development of multi-degree-of-freedom platforms. Methods: By using the research methodology, research substance, and inventive structure of the most recent representative patents of the multi-degree-of-freedom platform, the platform's operating principle and characteristics are demonstrated. Results: By contrasting multi-degree-of-freedom platforms in various application domains, the issues with current multi-degree-of-freedom platforms are enumerated, and the platforms' potential future development path and research topics are suggested. Conclusion: The growth of the aviation, medical, and military industries is aided by the development of multi-degree-of-freedom platforms, and the flexible distribution of high-precision parallel platforms has promising future development.

A Survey on Self-Supervised Learning: Algorithms, Applications, and Future Trends.

Deep supervised learning algorithms typically require a large volume of labeled data to achieve satisfactory performance. However, the process of collecting and labeling such data can be expensive and time-consuming. Self-supervised learning (SSL), a subset of unsupervised learning, aims to learn discriminative features from unlabeled data without relying on human-annotated labels. SSL has garnered significant attention recently, leading to the development of numerous related algorithms. However, there is a dearth of comprehensive studies that elucidate the connections and evolution of different SSL variants. This paper presents a review of diverse SSL methods, encompassing algorithmic aspects, application domains, three key trends, and open research questions. First, we provide a detailed introduction to the motivations behind most SSL algorithms and compare their commonalities and differences. Second, we explore representative applications of SSL in domains such as image processing, computer vision, and natural language processing. Lastly, we discuss the three primary trends observed in SSL research and highlight the open questions that remain.

Explainable machine learning models for predicting the acute toxicity of pesticides to sheepshead minnow (Cyprinodon variegatus)

A quantitative structure–activity relationship (QSAR) study was conducted on 313 pesticides to predict their acute toxicity to Sheepshead minnow (Cyprinodon variegatus) by using DRAGON descriptors. Essentials accounting for a reliable model were all considered carefully, giving full consideration to the OECD (Organization for Economic Co-operation and Development) principles for QSAR acceptability in regulation during the model construction and assessment process. Nine variables were selected through the forward stepwise regression method and used as inputs to construct both linear and nonlinear models. The obtained models were validated internally and externally. Generally, machine learning-based methods, namely support vector machine (SVM), random forest (RF), and projection pursuit regression (PPR), perform better than the multiple linear regression (MLR) model. The statistical results (R2 = 0.682–0.933, Q2LOO = 0.604–0.659, Q2F1 = 0.740–0.796, CCC = 0.861–0.882) of the developed models show that they are robust, reliable, reproducible, accurate and predictive. Comparatively, the RF model performs best, giving predictive correlation coefficient Q2 of 0.814, root mean squared error (RMSE) of 0.658 and mean absolute error (MAE) of 0.534 for the test set, respectively. The RF model (as well as SVM and PPR models) was visualized and explained by using the SHapley Additive explanation (SHAP) analysis to enhance its transparency and credibility. In addition, the applicability domain (AD) range of the RF model was characterized by the Williams plot and the tree manifold approximation and projection (TMAP) technology was utilized to illustrate similarity and diversity of the entire data space, to assist in the analysis of the outliers. Activity cliff detection was investigated by using Arithmetic Residuals in K-groups Analysis (ARKA) descriptors. It was found that none of the pesticides was identified as an activity cliff in the training set or a potential prediction cliff in the test set. Therefore, the RF model fulfills each OECD principle in regulation for QSAR models. The research in this work will aid in the in silico QSAR prediction of the acute toxicity to Sheepshead minnow (Cyprinodon variegatus) for untested and new toxic pesticides and can also be extended to other studies.

Innovative fabrication of eco-friendly bio-based foam from sugarcane bagasse and sodium alginate with enhanced properties and sustainable applications for plastic replacement

Foam materials possess notable features, including low density, high porosity, low thermal conductivity, and high impact resistance, making them extensively used in various sectors such as construction, transportation, water treatment, and packaging. However, the majority of commercial foam materials are derived from synthetic polymers based on fossil fuels, which raises concerns regarding health and ecology. In this work, we present a straightforward and scalable approach (physical cross-linking and common pressure drying) for fabricating a bio-based foam material using pulp fibers sourced from discarded sugarcane bagasse and sodium alginate extracted from seaweed. The resulting foam material exhibits low density (13.7–20.5 kg/m3), exceptional thermal insulation properties (thermal conductivity as low as 38.6 mW/mK), thermal stability (no deformation at 140 °C), and mechanical strength. After a basic silane modification, it also shows favorable water resistance (with a contact angle of 128°). Moreover, the foam material demonstrates remarkable degradation performance, with a degradation rate exceeding 93.8 % after 90 days of burial in soil. Therefore, this novel method offers a sustainable solution for eco-friendly foam production across various application domains.

Artificial Intelligence in Engineering Education: A Review of Pedagogical Innovations

This study critically evaluates the integration of artificial intelligence (AI) into engineering education, with a specific focus on its transformative role within the Digital-Intelligent Era. Drawing on General Systems Theory (GST), this research synthesizes insights from 82 empirical studies spanning 2011 to 2024, exploring AI's impact on the evolution of educational frameworks. The study identifies six key application domains of AI within engineering education, including intelligent tutoring systems, adaptive learning platforms, virtual laboratories, and personalized curriculum design. It highlights the synergy between AI and other emerging technologies, such as 5G, Cloud Computing, and Big Data, driving pedagogical innovation and enhancing the learning experience. Additionally, the paper addresses challenges related to the implementation of AI-based educational strategies, including infrastructure limitations, resistance to change, and equity concerns. Finally, it offers strategic solutions to overcome these challenges, fostering a more inclusive and effective educational environment.

Elucidating Molecular Mechanism and Chemical Space of Chalcones through Knowledge Graph and Machine Learning: A Combined Bio-Cheminformatic Pipeline

We developed a bio-cheminformatics method, exploring disease inhibition mechanisms using machine learning-enhanced quantitative structure-activity relationship (ML-QSAR) models and knowledge-driven neural networks. ML-QSAR models were developed using molecular fingerprint descriptors and the Random Forest algorithm to explore the chemical spaces of Chalcones inhibitors against diverse disease properties, including antifungal, anti-inflammatory, anticancer, antimicrobial, and antiviral effects. We generated and validated robust machine learning-based bioactivity prediction models (https://github.com/RatulChemoinformatics/QSAR) for the top genes. These models underwent ROC and applicability domain analysis, followed by molecular docking studies to elucidate the molecular mechanisms of the molecules. Through comprehensive neural network analysis, crucial genes such as AKT1, HSP90AA1, SRC, and STAT3 were identified. The PubChem fingerprint-based model revealed key descriptors: PubchemFP521 for AKT1, PubchemFP180 for SRC, PubchemFP633 for HSP90AA1, and PubchemFP145 and PubchemFP338 for STAT3, consistently contributing to bioactivity across targets. Notably, chalcone derivatives demonstrated significant bioactivity against target genes, with compound RA1 displaying a predictive pIC50 value of 5.76 against HSP90AA1 and strong binding affinities across other targets. Compounds RA5 to RA7 also exhibited high binding affinity scores comparable to or exceeding existing drugs. These findings emphasize the importance of knowledge-based neural network-based research for developing effective drugs against diverse disease properties. These interactions warrant further in vitro and in vivo investigations to elucidate their potential in rational drug design. The presented models provide valuable insights for inhibitor design and hold promise for drug development. Future research will prioritize investigating these molecules for mycobacterium tuberculosis, enhancing the comprehension of effectiveness in addressing infectious diseases.

Deep Learning Methods for De Novo Peptide Sequencing.

Protein tandem mass spectrometry data are most often interpreted by matching observed mass spectra to a protein database derived from the reference genome of the sample being analyzed. In many application domains, however, a relevant protein database is unavailable or incomplete, and in such settings de novo sequencing is required. Since the introduction of the DeepNovo algorithm in 2017, the field of de novo sequencing has been dominated by deep learning methods, which use large amounts of labeled mass spectrometry data to train multi-layer neural networks to translate from observed mass spectra to corresponding peptide sequences. Here, we describe these deep learning methods, outline procedures for evaluating their performance, and discuss the challenges in the field, both in terms of methods development and evaluation protocols.

Advancements in Computer Vision: A Comprehensive Survey of Image Processing and Interdisciplinary Applications

Computer vision and image processing are rapidly evolving fields with broad applications across numerous domains, including healthcare, autonomous driving, surveillance, and entertainment. These fields have transformed from simple data recording techniques into sophisticated systems that incorporate digital image processing, pattern recognition, machine learning, and computer graphics. This evolution has prompted interdisciplinary interest, pushed the technology’s boundaries and expanded its practical uses. This paper offers a comprehensive survey of recent advancements in computer vision, focusing on image processing and its applications across various fields. It delves into the theoretical foundations and technologies that make computer vision a valuable tool for interpreting images and videos, extracting relevant information, recognizing patterns, and understanding events. The ability of computer vision to analyze large datasets across multiple application domains makes it instrumental in tasks such as object identification, facial recognition, scene understanding, and even real-time action prediction. This versatility has established computer vision as a key driver of data-driven insights in both scientific and commercial sectors. The study categorizes computer vision into four main areas: image processing, object recognition, machine learning, and computer graphics. Each of these categories is essential to the functionality of modern computer vision systems. Image processing involves techniques for enhancing image quality and extracting important features. Object recognition and machine learning enable the identification of specific elements within images and allow systems to learn from large datasets, enhancing accuracy over time. Computer graphics, on the other hand, aid in visualizing and interpreting processed data. By offering insights into the latest techniques and evaluating their performance, this survey highlights the current state of computer vision while shedding light on future trends. Computer vision’s expanding utility across various fields underscores its critical role in driving interdisciplinary innovation and addressing complex challenges.

Optoelectronic metadevices.

Metasurfaces have introduced new opportunities in photonic design by offering unprecedented, nanoscale control over optical wavefronts. These artificially structured layers have largely been used to passively manipulate the flow of light by controlling its phase, amplitude, and polarization. However, they can also dynamically modulate these quantities and manipulate fundamental light absorption and emission processes. These valuable traits can extend their application domain to chipscale optoelectronics and conceptually new optical sources, displays, spatial light modulators, photodetectors, solar cells, and imaging systems. New opportunities and challenges have also emerged in the materials and device integration with existing technologies. This Review aims to consolidate the current research landscape and provide perspectives on metasurface capabilities specific to optoelectronic devices, giving new direction to future research and development efforts in academia and industry.