Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.

Towards the Intelligent Internet of Everything

147 vues

Publié le

In this presentation, Prof. Theo Lynn (DCU) was talking about observations on Multi-disciplinary Challenges in Intelligent Systems Research, at the RECAP consortium meeting in Dublin, Ireland on 06 November 2018.

Publié dans : Internet
  • Soyez le premier à commenter

  • Soyez le premier à aimer ceci

Towards the Intelligent Internet of Everything

  1. 1. Reliable Capacity Provisioning and Enhanced Remediation for Distributed Cloud Applications http://recap-project.eu recap2020 THIS PROJECT HAS RECEIVED FUNDING FROM THE EUROPEAN UNION’S HORIZON 2020 RESEARCH AND INNOVATION PROGRAMME UNDER GRANT AGREEMENT NUMBER 732667 Towards the Intelligent Internet of Everything Observations on Multi-disciplinary Challenges in Intelligent Systems Research Theo Lynn, Irish Centre for Cloud Computing and Commerce Coloquio de Doctorados 2018: Tecnología, Ciencia y Cultura: Una Visión Global 06 November 2018
  2. 2. 2 I got us an expert to help us evolve our thinking on the Internet Of Everything. Blah, blah, blah cognitive architectures Blah, blah, blah artificial intelligence Blah, blah, blah trust Its as if you’re he thinks he’s a technologist and philosopher all in one!!! Combining the Internet of Everything and Intelligent Systems in 50 minutes is not an easy thing!
  3. 3. Agenda • Conceptualising the Internet of Everything • Intelligent Systems ⁃Architecture Design Principles ⁃Reasoning and Information Processing • Some Observation on Research Challenges ⁃Ubiquitous Sensing ⁃Cognitive Architectures ⁃Infrastructure ⁃Trust • Q&A 3
  4. 4. The Internet of Everything presages a society where social structures and activities, to a greater or lesser extent, are organized around digital information networks that connect people, processes, things, data and social networks. Cisco, 2017
  5. 5. Cost, performance and form factor breakthroughs in sensors combined with advances in big data and cloud technologies are creating new value networks and economic opportunities Src: Dormon, 2014
  6. 6. By 2020, Gartner predicts that the world will contain more than 20 billion IoT devices, representing less than 1.4% of all physical objects worldwide 6 Src: Cisco, 2017
  7. 7. The value of the IoE to society is significant. It is estimated to generate up to US $19 trillion by 2022 • increased asset utilization and employee productivity, • improved supply chain and logistics, • optimized customer experience, and • accelerated innovation Public Sector $4.6tr Private Sector $14.4tr Src: Cisco, 2013
  8. 8. The Internet of Everything represents a significant research programme even before we make it smart….
  9. 9. 9 What do we mean by intelligent systems?
  10. 10. James Albus Intelligent systems are systems that act appropriately in an uncertain environment, where appropriate action is that which increases the probability of success, and success is the achievement of behavioral subgoals that support the system’s ultimate goal. Albus & Meystel, 1997
  11. 11. Albus & Meystel (1997) conceptualised intelligent systems as four functional elements –behaviour generation, sensory perception, world modelling and value judgment The planning and control of action designed to achieve behavioral goals through agents. The transformation of data from sensors into meaningful and useful representations of the world. • a) the computation of cost, risk, and benefit of actions and plans, • b) the estimation of the importance and value of objects, events, and situations, • the assessment of reliability of information, • the calculation of reward or punishment resulting from perceived states and events. • Uses sensory input to construct, update, and maintain a knowledge database. • Answers queries from behavior generation regarding the state of the world. • Simulates results of possible future plans. • Generates sensory expectations based on knowledge in the knowledge database.
  12. 12. These are supported by a knowledge database that provides a best estimate of the state of the world and the processes and relationships that effect events in the word. • The knowledge database contains (i) state variables, (ii) entity frames, (iii) event frames, (iv) rules and equations, (v) images, (vi) maps, and (vii) task knowledge. • The knowledge database has both long (static or slowly varying) and short (dynamic) memory. • Entities-of-attention are entities that have either been specified by the current task, or are particularly noteworthy entities observed in current memory input.
  13. 13. A communications systems manages the interactions between modules, agents, and nodes. Complexity through hierarchical layering and focussed attention.
  14. 14. Self-organization is a dynamical and adaptive process where systems acquire and maintain structure themselves, without external control De Wolf and Holvoet (2004). This implies the absence of external control or interference from outside the boundaries of the system.. Autonomy Self-organization is a process from dynamism towards order. Dynamical A self-organizing system must be capable of maintaining its organization autonomously in the presence of changes in its environment. It may generate different tasks but maintain the behavioral characteristics of its constituent parts. Adaptability or robustness with respect to changes An increase in order (or statistical complexity), through organization, is required from some form of semi-organized or random initial conditions to promote a specific function. Increase in Order
  15. 15. 15 Autonomic Computing and Self* Principles Components and systems continually seek opportunities to improve their own performance and efficiency. Self Optimisation Self-protection Automated configuration of components and systems follows high-level policies. Rest of system adjusts automatically and seamlessly. Self Configuration System automatically detects, diagnoses, and repairs localised software and hardware problems. Self-healing System automatically defends against malicious attacks or cascading failures. It uses early warning to anticipate and prevent system-wide failures.
  16. 16. 16 Architecture specifies inputs and outputs of each module and protocols for communication Multi-agent Systems Modules read and alter a shared memory of beliefs, goals and short term structures Functional processes that operate on structures including performance and learning mechanisms Modules communicate directly with each other No direct communication between modules Representation and organization of structures embedded in memories Distinct modules for different facets of an intelligent system Distinct modules for different facets of an intelligent system Short-term and long- term memories that store the agent’s beliefs, goals, and knowledge Architecture places no constraints on how each component operates A programming language to construct knowledge-based systems that embody the architecture’s assumptions Blackboard Systems Cognitive Architectures Architectural paradigms in intelligent systems
  17. 17. 17 Cognitive architectures reason about problems across different domains, develop insights, adapt to new situations and reflect on themselves.
  18. 18. Cognitive architectures make use of a wide variety of representation and information processing methods but are trending towards hybrid appoaches
  19. 19. Me! An Intelligent Internet of Everything as a system of systems that connects people, processes, things, data, and social networks and through intelligent systems proactively creates new value for individuals, organizations and society as a whole
  20. 20. Sensors play a vital role as an operational technology in IOT/IOE that gathers data to enable decision making. Sensors perform a wide variety of functions with a range of information utility Src: AMR, 2016 Src: Gemelli, 2017
  21. 21. Ambient intelligence assumes a digitally-infused environment that proactively, but sensibly, supports people in their daily lives (Ramos et al. 2008).
  22. 22. Ubiquitous sensing research requires a multi-disciplinary and inter- disciplinary approach Ubiquitous Sensing New materials and detection methods Greater understanding of the underlying bio-physical mechanisms, activities and events. Computational performance and interoperability Sensor Evaluation and Validation
  23. 23. 02 Adequate experimental validation and reproducibility of results Testing in diverse, challenging and realistic environments, real-world situations, more elaborate scenarios, and diverse tasks. Fuller technical data and access to software and data. Some cognitive architecture research priorities (Kotseruba & Tsotsos, 2016) 03 Human-like learning More robust and flexible learning mechanisms, knowledge transfer, and accumulation of knowledge without affecting prior learning. 04 Realistic Perception Advancements in active vision, localization and tracking, performance under noise and uncertainty, and the use of context information to improve detection and localization. 01 Autobiographic memory Episodic memory and lifelong memory. 05 Comparative evaluation of cognitive architectures Combination of (i) objective and extensive evaluation procedures and (ii) theoretical analysis, software testing techniques, benchmarking, subjective evaluation and challenges, to every aspect of the cognitive architecture and probing multiple abilities. 06 Natural Communications Advancements in verbal communications including knowledge bases for generating dialogues, robustness, detection of emotional response and intentions, personalized responses as well as performing and detecting other non-verbal human communications.
  24. 24. 25 What kind of research does a cloud research centre do on the Internet of Everything?
  25. 25. 26 Traditional cloud infrastructure is built on commoditised resources and is not optimised for high throughput/performance computing, heterogeneous processors and end-devices Simplified operational model for the cloud service provider, the developer and the end user through blueprint-as-a-service and constrained self-organisation. Novel scheduler architecture enables multiple scheduling logics and results in significantly higher task throughput, computational resource management and energy efficiency. Increased extensibility to work with new heterogeneous hardware through a plug & play service and new applications through blueprint- as-a-service. Supports on-premise, private clouds, public clouds, hybrid clouds and other federated computing environments. Greater scalability through novel self- organising self-managing approach.
  26. 26. Gateway Service Self Organizing Self Management System Plug & Play Service Blueprint Creator End User Services Catalogue Blueprint Catalogue Enterprise Cloud Operator Gateway Service UI Heterogeneous Resources New Hardware Deploy Service Service User Perspective Monitor Request to join CL-Resource Discover Resource Extract / Modify Blueprints Request Resource CL-Resources Deploy Blueprint Running Service Extract Blueprint Get Services Create Blueprints Get Status Resource Handler
  27. 27. 28 CloudLightning uses a combination of self- organisation, self- management, and separation of concerns to manage complexity in cloud infrastructure
  28. 28. 29 • Something is wrong on your end • Something is wrong on the Internet • Something is wrong on the other end
  29. 29. IOE / IOT EDGE CORE CLOUD DATA CENTRE RECAP Use Cases • Use Case A: Infrastructure and Network Management • Use Case B: Big Data Analytics Engine • Use Case C: Edge/Fog Computing for Smart Cities • Use Case D.1: Virtual Content Distribution Networks • Use Case D.2: Network Function Virtualization
  30. 30. IOE / IOT EDGE CORE CLOUD DATA CENTRE RECAP Use Cases • How many instances? Which hardware for which application • What data centre should we use? • Which parts to offload? Where to offload to? (When to) move parts around?
  31. 31. RECAP is a 3 year project that seeks to use advanced modelling and analytics to improve network and cloud deployment and remediating in IOT/IOE scenarios
  32. 32. Scale is a huge problem in the target use scenarios requiring new approaches to simulation. • Access to the data • Experiment model size • Network topology (Graph) • Infrastructure (Physical and Virtual) • Workload • Application • Speed and resource demand • Balance between accuracy and granularity
  33. 33. RECAP Optimisations 34 Decrease Cost Increase Revenue Reliability Competitive Necessity Provide superior products and services Achieve heightened market penetration Develop & deploy services quickly ICT infrastructure quality Digital service innovation Cost-efficient and agile Relationship infrastructure End-to-end service chain observability Anomaly detection and remediation Infrastructure and application reliability Improved clarity of costs Auto-scale infrastructure accurately Faster software releases Reduced investment Accurate planning and forecasting Reduced human ICT support
  34. 34. 35
  35. 35. There is a significant trade-off between consumer acceptance of sensing/surveillance and their exploitation through sensing/surveillance (Acquisti, 2008; Singh & Lyon, 2013). An interactive realm wherein every action and transaction generates information about itself Andrejevic, 2007 The ability to sense what customers will want next, knowing what they will ask before they request it. Franzak et al. 2001 The Culture The Digital Enclosure
  36. 36. What is trust? The intention to accept vulnerability based upon positive expectations of the intentions or behavior of another (Rousseau et al. 1998) Trust Benevolence Ability Integrity Mayer et al. 1995
  37. 37. Different interpersonal cues inform our trust behaviour. 3 2 1 Training, Title, Reputation, Code of Ethics Role 02 03Organisation Norms, Traditions, Practices, Semiotics Rules 01Gender, race, accent, attire Identity
  38. 38. Bio-chemistry also plays a role. Oxytocin increases trust among humans (Baumgartner et al. 2008) Mori, 1970
  39. 39. Our emotions towards things as they become to look more human- like is complicated Mori, 1970
  40. 40. Intelligent systems are based on input and training from humans and the environment can reflect our biases incl. race, gender and age (Caliskan et al. 2017).
  41. 41. The Internet of Everything introduces a wide range of new security risks and challenges including using connected end-points as attack vectors
  42. 42. Is trust in technology the same as trust in humans? Helpfulness Reliability and Predictability Functionality and Performance Benevolence Integrity Ability McKnight et al. 2011; Sollner et al. 2013
  43. 43. There are a lot of trust issues to be resolved in the Internet of Everything 1. Choice of law/jurisdiction 2. Data location and transfer to countries outside of the EEA 3. Data integrity and availability 4. Security of data 5. IP Issues • Copyright (incl. ownership of metadata) • Patents and trade secrets 7. Liability and indemnities 8. Acceptable use requirements 9. Service levels and performance across a complex service chain 10. Variation of contract terms across providers 14. Monitoring 15. Backup 16. Termination • Data / application preservation • Data transfer • Data deletion
  44. 44. Adding intelligence might manage technical complexity but introduce additional trust complexity
  45. 45. What does assurance and accountability mean in an Intelligent Internet of Everything?
  46. 46. Martin Rees I don’t but worries about AI expanding and taking over the universe. Our evolution has required two things: intelligence and a certain degree of aggression. There is no reason why machines that are not intelligent should also be aggressive Astronomer Royal, Emeritus Professor of Cosmology and Astrophysics at the University of Cambridge
  47. 47. Vijay Saraswat, What does the notion of ethics mean for a machine that does not care whether it or those around it continue to exist, that cannot feel, that cannot suffer, that does not know what fundamental rights are? Chief Scientist for IBM Compliance Solutions
  48. 48. Joseph Stiglitz Which is the easier way to make a buck: figuring out a better way to exploit somebody, or making a better product? With the new AI, it looks like the answer is finding a better way to exploit somebody. Nobel Laureate, Former Chief Economist, World Bank
  49. 49. 50 cuando?
  50. 50. We are decades away from having the building blocks for an Intelligent Internet of Everything More than 10 Years • Biotech – cultured or artificial tissue • Artificial General Intelligence • 4D Printing • Human Augmentation • Brain-Computer Interface
  51. 51. The timing for research is just right The Goldilocks Dilemma and Internet of Everything Research
  52. 52. THANK YOU http://recap-project.eu recap2020 RECAP Project ■ H2020 ■ Grant Agreement #732667 Call: H2020-ICT-2016-2017 ■ Topic: ICT-06-2016 THIS PROJECT HAS RECEIVED FUNDING FROM THE EUROPEAN UNION’S HORIZON 2020 RESEARCH AND INNOVATION PROGRAMME UNDER GRANT AGREEMENT NUMBER 732667 Theo Lynn theo.lynn@dcu.ie