**4.2 Cognitive architectures**

*Technology, Science and Culture - A Global Vision*

advanced functionalities into sensors but also drive down costs, thereby increasing adoption and proliferation of sensors [36]. Similarly, nanotechnology is contributing to the development of electrochemical, optical, and magnetic resonance biosensors that are transforming our understanding of biology and treatment of disease including novel drug discovery for this purpose [37, 38]. As well as *in vivo* sensors, increasingly sophisticated sensors are providing multimodal sensing to not only allow machines to see and hear but also touch [39], taste [40], and smell [41] like humans. Advances in computer processing, networking, and miniaturization are introducing a new generation of bio-inspired mobile, wearable, and embedded sensors. These bio-inspired sensors are enabling machines to detect, alert, and prevent harmful events such as chemical threats in an unobtrusive fashion previously the

*The ambient-intelligence vision from an artificial intelligence perspective [34].*

Despite the progress in developing increasingly novel and sophisticated sensors, Paulovich et al. [36] highlight the need for greater research into new materials and detection methods for wearable and implantable devices and specifically device engineering for both producing and commercializing low-cost, robust sensing units. Like all interdisciplinary research areas, the success of sensors and biosensing requires not only greater understanding of the technologies involved but also the underlying biophysical mechanisms, activities, and events inspiring sensors and specifically biosensor development. Furthermore, before intelligent systems can be relied upon particularly in high criticality situations or *in vivo*, concerns relating to predictability, reproducibility, and sensor drift and calibration, amongst others

**60**

**Figure 1.**

domain of science fiction [42].

must be addressed over a sustained period [36].

Ambient intelligence and the wider Intelligent Internet of Everything assume that once our sensing network captures data, we will have the systems in place to analyze the various sensing inputs to interpret, analyze, plan, act, and otherwise interact with other systems and humans [34]. Despite significant advancements, the state-of-the-art in cognitive architectures remains at a relatively early stage relative to match the 3000+ cognitive abilities of humans [43]. Both Ramos et al. [34] and Kotseruba and Tsotsos [14] emphasize the need for advancements in approaches to model human cognition and increase the range of supported cognitive abilities. To this end, Kotseruba and Tsotsos [14] identify eight areas for further research based on their survey of extant research on cognitive architectures:


## **4.3 Infrastructure**

Conventional cloud architectures allocate functionality against a very different set of design principles than the emerging Intelligent Internet of Everything. Decisions are made within the boundaries of a cloud and its infrastructure. The emergence of the Internet of Everything requires the allocation and/or optimization of resources and activities along a Cloud-to-Things (C2T) continuum and to accommodate computation at the edge but also using intermediary data management and processing capabilities or fog computing as it is commonly known. Latency requirements, network bandwidth constraints, device profiles, uninterrupted service

provision with intermittent connectivity to the Cloud, cyber-physical devices, and security are just some of the design and research challenges that intelligent systems and the Internet of Everything will face [44]. Based on the scale, device (thing) heterogeneity, dynamism of the Internet of Everything, and existing approaches for resource provisioning and remediation are inadequate. As this complexity increases, this need will become increasingly acute.

The cloud is an essential component of future large-scale intelligent systems and the Internet of Everything. However, it requires fundamental changes to how they and their underlying infrastructure is designed and managed. Cloud computing data centers today largely leverage homogeneous hardware and software platforms to support cost-effective high-density scale-out strategies. The advantages of this approach include uniformity in system development, programming practices, and overall system capability, resulting in cost benefits to the cloud service provider [45]. There are significant disadvantages too not least energy inefficiencies and sub-optimal performance for specific use cases. Intelligent systems require significant data management support both for memory-storage and real-time processing [14, 36]. While the cloud has near-infinite low cost storage, intelligent systems will require high throughout communications. Similarly, existing commodity processors may not be suitable for the information processing tasks and techniques used in the cloud. Specialist coprocessor architectures, with relatively positive computation/ power consumption ratios, are emerging which are better equipped to deal with the intelligent methods and techniques including GPUs, many integrated cores (MICs), and data flow engines (DFEs); however, their use in a cloud context is a specialist and nice market.

As in the wider domain of intelligent systems, concepts such as emergence and self-principles are being explored as methods to address complexity in the underlying networks and infrastructure that supports today's Internet and the future Internet of Everything. For example, Östberg et al. [46] seek to specifically address the optimization of network and cloud resources in Internet of Things scenarios through more intelligent instantiation of services close to the end users that require them across the C2T continuum. Given the dynamism and nature of the Internet of Everything and to address greater variability in Quality of Service (QoS) levels, future infrastructure will need to be able to pre-emptively take reconfiguration and remediation actions in a fully autonomic fashion. They argue that this can be achieved through the concerted activation of (i) the prediction of the evolution of workload and application performance, (ii) the simulation of different deployments across the C2T continuum, (iii) the optimization of the deployment given the output of historic and real-time analysis, and (iv) the relocation of services and application components to achieve the required QoS. Similarly, Xiong et al. [47] propose a novel architecture to manage heterogeneous resources (including new processor architectures) and to improve service delivery in clouds based on a loosely coupled, hierarchical, self-adapting management model, deployed across multiple layers. This project is noteworthy in the context of intelligent systems as it specifically includes mechanisms to identify and provision appropriate resources (including specialist processors) to support the process parallelism associated with high performance services such as those required by intelligent systems.

#### **4.4 Trust, privacy, and security**

Trust is commonly defined as "the intention to accept vulnerability based upon positive expectations of the intentions or behavior of another" ([48], p. 395). Trust decisions are typically made based on an assessment of (i) ability, (ii) benevolence, and (iii) integrity [49]. These have been transposed to

**63**

assurance and accountability.

*Toward the Intelligent Internet of Everything: Observations on Multidisciplinary Challenges…*

the IT domain as (i) functionality and performance (the capability of a technology), helpfulness (access and support when using a technology), and reliability and predictability (consistency and dependability) [50, 51]. There is a large established literature that suggests that trust, and the lack thereof, is a major driver, and barrier, of information and communication technology adoption. This is particularly the case at the early stages of adoption and interaction with a new technology or service and is further complicated by environmental uncertainties as well as the risks posed by the potential of the malfunctioning of a product or service and the acts of malicious third parties [52–55]. Intelligent systems and the

As well as the technological logistics of data management in intelligent systems, it is foreseeable that end users will face similar concerns to cloud computing relating to the location, integrity, portability, security, and privacy of their data both as used by intelligent systems and the wider Internet of Everything. Indeed the underlying technologies we have discussed—ubiquitous sensing, cognitive architectures, and self-organizing, self-managing infrastructure—are all likely to exacerbate data

Ubiquitous sensing can lead to the capture and storage of data indiscriminately and indeed with the permission of consumers [56]. Consumers may accept increased surveillance and intrusions into their privacy for a variety of reasons, not least the benefits outweighing the costs and perceptions of digital inevitability and transformation [57]. Some commentators have argued that consumers may indeed be technologically unconscious and simply unaware of what happens to their data and the implications for their privacy, identity, and well-being [58]. Indeed in the context of ambient intelligence data on third parties may be captured without their knowledge or express permission at all. There is a significant trade-off between data privacy and data utility and consumer acceptance of sensing/surveillance and their exploitation through sensing/surveillance [57, 59]. Many commentators have suggested that ubiquitous sensing/surveillance can erode reasonable expectations of privacy for society as a whole and result in mental health issues, pernicious or the perception of pernicious surveillance by others including organizations and government as well as changing how we think about accountability and identity [56]. Cognitive architectures similarly present issues around trust, privacy, and security. Following on from McKnight et al. [50] and Söllner et al. [51], Kotseruba and Tsotsos [14] suggest a need for greater and more robust evaluation, validation, and reproducibility before the performance, reliability, and predictability before they can be trusted. The reasoning and information processing techniques used in intelligent systems are based on human knowledge, which, as a result, include human biases including race, gender, and age [60]. Specific techniques based on emergent paradigms present other issues relating to transparency and as a result

Trust issues relating to infrastructure and particularly the cloud are well documented. Governance issues, data loss and leakage, and shared technology vulnerabilities are amongst the top threats to cloud computing identified by independent international organizations [61]. These are likely to present themselves in the context of the Intelligent Internet of Everything and raise complex ethical, legal, and regulatory questions. Who owns data and metadata generated by the interaction of the various entities in the Internet of Everything? How will data integrity and availability be managed? What is acceptable use of Internet of Everything data? Who will regulate it, if at all? These issues will be exacerbated by an extremely complex, most likely cross-border, and chain of service provision. Similarly, securing the Internet of Everything will likely be a significant challenge in itself. The envisioned proliferation of hyper-connected entities is and will introduce a wide range of new

*DOI: http://dx.doi.org/10.5772/intechopen.83691*

Internet of Everything are not immune from these risks.

privacy risks and concerns rather than ameliorate them.

#### *Toward the Intelligent Internet of Everything: Observations on Multidisciplinary Challenges… DOI: http://dx.doi.org/10.5772/intechopen.83691*

the IT domain as (i) functionality and performance (the capability of a technology), helpfulness (access and support when using a technology), and reliability and predictability (consistency and dependability) [50, 51]. There is a large established literature that suggests that trust, and the lack thereof, is a major driver, and barrier, of information and communication technology adoption. This is particularly the case at the early stages of adoption and interaction with a new technology or service and is further complicated by environmental uncertainties as well as the risks posed by the potential of the malfunctioning of a product or service and the acts of malicious third parties [52–55]. Intelligent systems and the Internet of Everything are not immune from these risks.

As well as the technological logistics of data management in intelligent systems, it is foreseeable that end users will face similar concerns to cloud computing relating to the location, integrity, portability, security, and privacy of their data both as used by intelligent systems and the wider Internet of Everything. Indeed the underlying technologies we have discussed—ubiquitous sensing, cognitive architectures, and self-organizing, self-managing infrastructure—are all likely to exacerbate data privacy risks and concerns rather than ameliorate them.

Ubiquitous sensing can lead to the capture and storage of data indiscriminately and indeed with the permission of consumers [56]. Consumers may accept increased surveillance and intrusions into their privacy for a variety of reasons, not least the benefits outweighing the costs and perceptions of digital inevitability and transformation [57]. Some commentators have argued that consumers may indeed be technologically unconscious and simply unaware of what happens to their data and the implications for their privacy, identity, and well-being [58]. Indeed in the context of ambient intelligence data on third parties may be captured without their knowledge or express permission at all. There is a significant trade-off between data privacy and data utility and consumer acceptance of sensing/surveillance and their exploitation through sensing/surveillance [57, 59]. Many commentators have suggested that ubiquitous sensing/surveillance can erode reasonable expectations of privacy for society as a whole and result in mental health issues, pernicious or the perception of pernicious surveillance by others including organizations and government as well as changing how we think about accountability and identity [56].

Cognitive architectures similarly present issues around trust, privacy, and security. Following on from McKnight et al. [50] and Söllner et al. [51], Kotseruba and Tsotsos [14] suggest a need for greater and more robust evaluation, validation, and reproducibility before the performance, reliability, and predictability before they can be trusted. The reasoning and information processing techniques used in intelligent systems are based on human knowledge, which, as a result, include human biases including race, gender, and age [60]. Specific techniques based on emergent paradigms present other issues relating to transparency and as a result assurance and accountability.

Trust issues relating to infrastructure and particularly the cloud are well documented. Governance issues, data loss and leakage, and shared technology vulnerabilities are amongst the top threats to cloud computing identified by independent international organizations [61]. These are likely to present themselves in the context of the Intelligent Internet of Everything and raise complex ethical, legal, and regulatory questions. Who owns data and metadata generated by the interaction of the various entities in the Internet of Everything? How will data integrity and availability be managed? What is acceptable use of Internet of Everything data? Who will regulate it, if at all? These issues will be exacerbated by an extremely complex, most likely cross-border, and chain of service provision. Similarly, securing the Internet of Everything will likely be a significant challenge in itself. The envisioned proliferation of hyper-connected entities is and will introduce a wide range of new

*Technology, Science and Culture - A Global Vision*

increases, this need will become increasingly acute.

provision with intermittent connectivity to the Cloud, cyber-physical devices, and security are just some of the design and research challenges that intelligent systems and the Internet of Everything will face [44]. Based on the scale, device (thing) heterogeneity, dynamism of the Internet of Everything, and existing approaches for resource provisioning and remediation are inadequate. As this complexity

The cloud is an essential component of future large-scale intelligent systems and the Internet of Everything. However, it requires fundamental changes to how they and their underlying infrastructure is designed and managed. Cloud computing data centers today largely leverage homogeneous hardware and software platforms to support cost-effective high-density scale-out strategies. The advantages of this approach include uniformity in system development, programming practices, and overall system capability, resulting in cost benefits to the cloud service provider [45]. There are significant disadvantages too not least energy inefficiencies and sub-optimal performance for specific use cases. Intelligent systems require significant data management support both for memory-storage and real-time processing [14, 36]. While the cloud has near-infinite low cost storage, intelligent systems will require high throughout communications. Similarly, existing commodity processors may not be suitable for the information processing tasks and techniques used in the cloud. Specialist coprocessor architectures, with relatively positive computation/ power consumption ratios, are emerging which are better equipped to deal with the intelligent methods and techniques including GPUs, many integrated cores (MICs), and data flow engines (DFEs); however, their use in a cloud context is a specialist

As in the wider domain of intelligent systems, concepts such as emergence and self-principles are being explored as methods to address complexity in the underlying networks and infrastructure that supports today's Internet and the future Internet of Everything. For example, Östberg et al. [46] seek to specifically address the optimization of network and cloud resources in Internet of Things scenarios through more intelligent instantiation of services close to the end users that require them across the C2T continuum. Given the dynamism and nature of the Internet of Everything and to address greater variability in Quality of Service (QoS) levels, future infrastructure will need to be able to pre-emptively take reconfiguration and remediation actions in a fully autonomic fashion. They argue that this can be achieved through the concerted activation of (i) the prediction of the evolution of workload and application performance, (ii) the simulation of different deployments across the C2T continuum, (iii) the optimization of the deployment given the output of historic and real-time analysis, and (iv) the relocation of services and application components to achieve the required QoS. Similarly, Xiong et al. [47] propose a novel architecture to manage heterogeneous resources (including new processor architectures) and to improve service delivery in clouds based on a loosely coupled, hierarchical, self-adapting management model, deployed across multiple layers. This project is noteworthy in the context of intelligent systems as it specifically includes mechanisms to identify and provision appropriate resources (including specialist processors) to support the process parallelism associated with

high performance services such as those required by intelligent systems.

upon positive expectations of the intentions or behavior of another"

Trust is commonly defined as "the intention to accept vulnerability based

([48], p. 395). Trust decisions are typically made based on an assessment of (i) ability, (ii) benevolence, and (iii) integrity [49]. These have been transposed to

**62**

**4.4 Trust, privacy, and security**

and nice market.

security risks and challenges for individuals but also organizations where connected end-points are used as attack vectors for distributed denial of service attacks. Together, all these factors will dilute assurance, accountability, and transparency with regards to data privacy, protection, and security unless addressed.
