Understanding concept computing
2 likes3,268 views
The keynote presentation on 'concept computing' discusses a paradigm shift from information-centric to knowledge-driven computing, emphasizing its six defining attributes that enable smarter software functionality and user experiences. It outlines how this semantic, model-driven approach impacts enterprise and consumer markets by offering a unified environment for managing and executing various models, thus enhancing system adaptability and reducing complexity. The presentation predicts that concept computing will revolutionize how systems understand and process information, demanding new forms of user interaction and integration.
Understanding concept computing
- 1. #ConceptComputing: Where Smart WORK & PLAY SPACES go mainstream Mills Davis Project10X mdavis@project10x.com 1-202-667-6400 This is a keynote presentation first given at the Association for the Advancement of Artificial Intelligence 2012 conference (AAAI 12) in Toronto, Canada. It discusses the vision and principles of "concept computing;" and how this paradigm shift impacts different aspects of software functionality and value.
- 3. This is the entrepreneur’s cautionary “forward looking statements” slide. I prefer to show it as a road sign.
- 4. This triptych is from a hypothetical comic strip entitled “Roy Lichtenstein meets an Internet of Things” (1) A young woman at the left pines over which approach and investment to make to realize her dreams. (2) The scene in the middle depicts difficulties we currently experience trying to share meanings and link knowledge models together so that both people and machines can reason with it. (3) The woman to the right depicts consumer expectations in the near future, when almost everything we encounter has some intelligence and communicates spontaneously -- we’ll be very angry if the computer doesn’t understand what we mean. Roy Lichtenstein meets The Internet of things Source: Project10X Setting the stage...
- 5. Topics • What is “concept computing”? • How will concept computing impact software functionality and value? Topics in this talk address concept computing: what it is, how it will impact us, and where is it already happening.
- 6. Let’s travel the yellow brick road. First the vision of concept computing; then discuss the impacts; then I’ll précis a company that’s already taking concept computing into the mainstream.
- 7. What is concept computing? •paradigm shift from information-centric to knowledge-driven patterns of computing. •Spectrum of knowledge representation, from search to knowing. •Synthesis of AI, semantic, model-driven, mobile, and User interface technologies. •solution Architecture where every aspect of computing is semantic and directly model-driven. •Development methodology where Every stage of the solution lifecycle becomes semantic, model-driven & super-productive. •new domain where value multiplies. Here are six defining attributes of concept computing.
- 8. what do we mean by concept? •A concept is semantic model. •The meaning of a concept derives from its network of relationships to other concepts. •The basic idea is to model concepts, relationships & contexts separately from the supporting IT systems and then to compute with these knowledge models. Source: Project10X
- 9. Source: Project10X #Conceptcomputing will drive the next internet. It’s smart computing -- for your mom! Concept computing will drive the next internet. It’s mainstream-ready, semantic, model-driven, activity-context aware technology for both consumer and enterprise markets. But, most important, concept computing is something that broad audiences can understand.
- 10. Source: Project10X The basic IT stack has user interface, application, information, and infrastructure layers. The internet has overlaid this with point and click network access to subjects, services, and things. Concept computing takes the next step. A concept plane interconnects and enables reasoning across all layers of the IT and internet stacks. Concept computing represents and processes knowledge about domains, user interface, application functionality, processes, information, and infrastructure separately from the underlying IT systems and other artifacts so that both people and computers can interpret concepts and put this knowledge to work.
- 11. Every aspect of the solution is model-driven, activity context-aware, and semantic. How is this different? Historically, lots of things have been modeled. But, modeling has only seemed cost-effective for individual aspects of software applications. Going back to the beginning of IT, there was only an application program. It was a deck of cards that gave instructions to a computer. It was low-level code. Over the decades, we began model knowledge about some things separately and take this functionality out of the application, so that multiple programs could share it. The sequence was something like this: operating systems, then data, workflow, rules, services, and goals. As modeling evolved, different kinds of concepts required separate tools to model them. With different kinds of modeling tools came different formalisms and standards. For example for: data schemas, decisions using business rules, processes flow-charted with BPMN, services accessed through APIs. Different formalisms and standards result in tools that don’t know about each other and don’t share semantics. That’s a problem when you want to combine multiple types of models in an application. It gets complicated. Often you are obliged to write some code. Other times, you import or export models into other tools, which adds a layer of complexity. With concept computing this ceases to be a problem. Concept computing provides a unified environment for creating, managing, and executing all types of models. Further, there is new hardware designed for concept computing at scale.
- 12. With concept computing, the model is the design, is the documentation, is the application, is the user interface. 1 When the model is the design, the solution self-documents. It’s just another way to express the model. 2 When the model is the application, the model executes at every stage of development. 3 When the application logs every everything that happens, the model can explain the rationale for every decision and action taken. 4 When the model drives the user interface and system behaviors, if you change devices, channels or the underlying model itself, then system behaviors update automatically. 5 Here is what’s different: — You don’t write program code. — You don’t draw flow charts in swim lanes either. — You don’t compose a waterfall of documents that translate requirements to designs to specifications to code and so on. — Business logic is packaged in knowledge models, and delivered as knowledge-as-a-service, where it can be reused by external applications. — One interconnected knowledge model directs activities and decisions dynamically towards the goal. — All system knowledge updates quickly, without your having to rebuild databases or compile new program code. — By the way, under the hood it’s all RDF & RDF/S.
- 13. #conceptcomputing reveals a new domain where value multiplies! A previous slide showed a concept plane being abstracted from the IT stack and the Internet network. This doesn’t tell the whole story. When every aspect of the solution is semantic and model driven, then a new world of value opens up. A vista or landscape in which innovations build on each other, and amplify in value.
- 14. Let’s pause for a moment to consider IBM’s vision of concept computing. Notice the three diagrams to the right. The first says we are entering the era of smart systems. The second identifies four technologies that will have a 1000X impact on capability and performance. The third shows a concept diagram for IBM’s WATSON.
- 15. Concept computing is a paradigm shift. This means it does things differently and enables capabilities that can be game changing. Concept computing impacts every aspect of software functionality. In the following slides we will discuss five areas.
- 16. Concept computing enables making user experience simpler, smarter and more helpful.
- 17. Source:Project10X Meanwhile, remember that consumer expectations are changing. Before long we’ll all be very angry if our computing devices don’t understand concepts.
- 18. Mobility is all about new user experience. Make no mistake about this. Mobile internet user experience demands computers that understand concepts. One illustration of this is Apple’s SIRI. What happens when semantic models enable computers to understand concepts? Computers that understand concepts are systems that know. That is, they are more than electronic pencils, more than calculators, and more than search appliances that retrieve information. Think about it. Systems that know can advise you. They can help you. They can simplify complex, knowledge-intensive tasks. They can adapt and optimize their behavior when events happen and something changes. But that’s not all. They can become virtual assistants that get something done for you.
- 19. Source:Project10X Go girl! Enterprises and governments too are getting this message. They’re asking for a SIRI behind the firewall -- They want a smart agent that communicates through an elegantly simple tablet style interface. Under the hood they want deep processing that knows everything about the domain, and task at hand, communicates in context, and helps you get things done.
- 20. Semantic and model-driven user interface design allows implementation of different types of “smarter” user experience. The progression is from fixed tools, to appliances, to advisors, to virtual assistants that can complete tasks, to expert agents.
- 21. If you want to connect and integrate information, the first thing you have to do is integrate what you know about it. Semantic web standards are gaining traction as a way of describing different data sources, structures and metadata so that they can be linked together. Concept computing goes further to put linked data and metadata to work.
- 22. Concept computing spans a comprehensive and expressive spectrum of knowledge representation (KR). More expressive KR powers greater reasoning capability. This figure shows a spectrum of executable knowledge representation and reasoning capabilities. As the rigor and expressive power of the semantics and knowledge representation increases, so does the value of the reasoning capacity it enables. From bottom-to-top, the amount, kinds, and complexity, and expressive power knowledge representation increases. From left-to-right, reasoning capabilities advance from: (a) Information recovery based on linguistic and statistical methods, to (b) Discovery of unexpected relevant information and associations through mining, to (c) Intelligence based on correlation of data sources, connecting the dots, and putting information into context, to (d) Question answering ranging from simple factoids to complex decision- support, to (e) Smart behaviors including robust adaptive and autonomous action. Moving from lower left to upper right, the diagram depicts a spectrum of progressively more capable forms of knowledge representation together with standards and formalisms used to express metadata, associations, models, contexts, and modes of reasoning. More expressive forms of metadata and semantic modeling encompass the simpler forms, and extend their capabilities. In the following topics, we discuss different forms of knowledge representation, then the types of reasoning capabilities they enable.
- 23. I like this slide from an Infosys manifesto about next generation work and play spaces. It says: “The real data revolution is in business structure and processes and how they use information for decision-making.” Big data is an investigation and a conversation with data. If you don’t know what answers you are seeking, or what you would do with them if you found them, then examining very large data sets isn’t going to help that much.
- 24. I sum it up this way: Concept computing puts semantic data to work. Concept computing uses semantic models to link sources; connect knowledge and data; enhance context; and, most important, integrate data, decisions, and actions. Semantic models encompass: • goal-oriented activities to perform • pre-and post-conditions for these activities • decisions required to take action; • rules and conditions to be met for choosing • data and calculations required. All model(s) are managed in one environment. Models are compact and integrated. For example, business rules always appear in context of their use. Schemas, ontologies, models, and business logic can be imported, exported, or updated using open standards. Concept computing can import linked data and ontologies in RDF/OWL and connect these to analytic, decision, and process models. Concept computing can combine natural language understanding with semantic models to extract and apply knowledge and information from unstructured sources.
- 25. Concept computing is about goal-oriented processes. Semantics are about meanings. Pragmatics go further. Pragmatics encompass goals, intentions and purposes. When pragmatic models power everything, you get to think differently. New capabilities and solution concepts become practical.
- 26. This diagram shows how concept computing impacts a spectrum of process types. 1. Fixed transaction processes follow a preset procedural sequence. Straight-through-processes are like this. So are simple workflows and instruction sequences. Trend is to use concept computing (semantic model driven) approaches when transaction systems need to be connected across boundaries. 2. Dynamic case management systems process events and rules in order to determine the specific sequence of steps to follow to reach a goal in this particular case. Modeling the potential variations can be complicate (for example, like a phone tree), or relatively elegant (like a GPS system) depending on how the process gets modeled. Trend is to use goal oriented, event driven concept computing approaches for administrative, investigative, and customer facing processes that are complex and knowledge intensive. Processes are compact and elegant. They adapt and self-optimize when events happen, exceptions occur, and needs change. 3. Emergent projects (like dynamic cases) have an underlying goal-oriented methodology (process model). However, they address problems for which not conditions can be pre-defined. Events can occur, which demand definition of a new task, methodology and deliverable outcome. The emergent process model evolves (learns) as well as adapts and self-optimizes.
- 27. Source:Project10X+BeInformed Imagine a multi-benefits solution that puts the user in the center of the action: •"A system knowledgebase integrates all legislation, regulation, and policies needed to guide the administrative process. •"A core business process pattern defines common high-level functions. •"These might include to inform, advise, apply for benefits, answer questions, decide eligibility, track status of cases, resolve exceptions, explain decisions, and communicate actions taken. •"Specific requirements of individual benefits programs are modeled as specializations. •"Every exception is just another business rule. •"Meanwhile, the user experiences a single interface where s/he can access information, advice, and obtain services for all benefit programs. •"Similarly, the caseworker has only to deal with actions actually needed for the specific case. •"Dynamic case management can reduce clicks and keystrokes required by a factor of ten.
- 28. Concept computing processes are goal-oriented, event-driven, and context-aware. Goal-oriented processes adapt, self-configure, and optimize when events happen, exceptions occur, or needs change. Like a GPS navigation system, the process interprets events and computes the next best action based on the current context, system knowledge, and content of the case. A dynamic activity plan continuously tracks and updates the status of actions taken in the system. No difference exists between straight through processing (STP) and exception handling. What can be automated is. What can’t, isn’t. It’s still the same process.
- 29. Concept computing empowers deeper reasoning that can deal with questions involving more than just logic. Value is the foundation of meaning. It is the measure of the worth or desirability (positive or negative) of something, and of how well something conforms to its concept or intension. Value formation and value-based reasoning are fundamental to all areas of human endeavor. Theories embody values. The axiom of value is based on “concept fulfillment.” Most areas of human reasoning require application of multiple theories; resolution of conflicts, uncertainties, and competing values; and analysis of trade-offs. For example, questions of guilt or innocence require judgment of far more than logical truth or falsity. Axiology is the branch of philosophy that studies value and value theory. Things like honesty, truthfulness, objectiveness, novelty, originality, “progress,” people satisfaction, etc. The word ‘axiology’, derived from two Greek roots ‘axios’ (worth or value) and ‘logos’ (logic or theory), means the theory of value, and concerns the process of understanding values and valuation.
- 30. Source:Project10X Is computing with knowledge only for niche applications? Concept computing overcomes difficulties of previous expert system, modeling, and model-driven technologies. Expert systems have been mainstays of AI and business rule driven applications for decades. But, there have been issues. The usual concerns cited are technical complexity, brittleness of the knowledge models, poor scalability of the solutions, difficulty to maintain, and lack of fit with mainstream programming paradigms. These concerns are no longer warranted. Concept computing is now robust, flexible, scalable, dynamic, and mainstream ready.
- 31. With concept computing, run-time semantic models provide a web of knowledge about infrastructure, information, application process, user interface, system behavior, and other domains. A concept plane cross-cuts and connects resources within and across each layer of the IT stack. It can be queried, interpreted and reasoned over both by people and machines. Semantic model-driven infrastructure opens a different dimension for architecture and development allowing cost-effective, sustainable solutions to problems of scale, complexity, connectedness, mobility, context, security, and interoperability.
- 32. This diagram shows the direction where infrastructure is headed. Infrastructure gets smarter. The trend depicted here is from fixed to semantic to autonomic, to autonomous components, systems, and ecosystems.
- 33. Concept Computing technologies can solve problems of scale, complexity, function, security, performance & agility. Key drivers of global economic expansion in the coming decade include the build-out of broadband telecommunications, mobility, and the deployment of intelligent services across this infrastructure (Mobile internet of subjects, things, and services). IT has reached the limits of what it can do with stacks, object orientation, metadata madness, fixed knowledge embedded in code (with no run-time learning), and architected development versus emergent solutions. Concept computing impacts infrastructure because it can enable overcoming problems of integration, interoperability, parallelism, mobility, ubiquity/pervasiveness, scale, complexity, speed, power, cost, performance, autonomics, automation, intelligence, identity, security, and ease of programming.
- 34. It turns out that media growth is a key driver for mobile growth as well as a key driver for growth of bandwidth and capacity.
- 35. Smart devices are transforming the internet. Both consumer and enterprise markets.
- 36. It’s not just smartphones and tablets that will play a role. Ambient displays, reality browsing, augmented reality all are coming into the conversation.
- 37. #ConceptComputing = supercompuiting goes mainstream Concept computing demands big think. Energy consumption is being worked on as an issue. Memory is becoming a non-issue. Meanwhile, super computing will be everywhere. Supercomputing at the edge. In smart devices. In the cloud. Intel. Nvidia. Cray. IBM. And more. All have supercomputing roadmaps and market plans.
- 38. Security challenges are shifting from: • Predictable, slowly-evolvable threat & risk models, interaction scenarios and behavior patterns to unpredictable and highly-dynamic ones; • Platform monopolies to massively distributed systems exhibiting unprecedented levels of sw/hw platform heterogeneity; • Device-and infrastructure-centric security models towards user-context and information-centric ones; • Predefined to opportunistic interactions with unknown parties in open, inherently insecure environments; and • Limited and fragmented data to unparalleled level of personal information richness and precision collected/processed/stored and communicated. Our expectation is for secure, trustworthy pervasive environments where: • Users control which data is being collected and the mannerin which it is aggregated, processed, stored and distributed; • Information is disclosed only to authorized parties and used for authorized tasks only; • Individuals are always sure with whom they are interacting; • Users are surrounded by millions of invisible, data collecting nano-devices building a huge, complex and dynamic system an omnipresent life-recorder; and • Data are captured continuously with unprecedented precision and completeness, both inside and outside us.
- 39. Concept computing, as with object oriented software before it, requires a new methodology, new tooling, and a new definition of “integrated development environment.” The new methodology for building solutions is knowledge-driven, not document centric. Every aspect of a process and every stage of the solution life cycle becomes model-driven and semantic. •"By aspect of a solution I include: user interaction, data, decisions, processes, and infrastructure. Everything. •"By stage of its life cycle I include: development, operations and ongoing evolution of the solution.
- 40. Source:Project10X With concept computing one of the biggest changes is “who does what?” Business users, subject matter experts, and IT specialists are all involved. Business users shape the business logic. Subject matter experts shape domain knowledge. And IT professionals shape the infrastructure and its operations. Each has the tools and user interface it needs. Concept computing handles all kinds of modeling in one environment. No more separate modeling tools and file formats.
- 41. Source:Project10X Concept computing enables everyone to model. Concept computing lets users express ideas in ways they find natural. The computer learns how to makes sense of it. There are multiple choices of user-friendly modeling methods. Examples include: graphical modeling, forms, spreadsheet style tabular modeling, and writing in controlled natural language.
- 42. Development using concept computing practices is fast and lean. Business users, subject matter experts, and IT specialists all participate in development and are involved throughout. Development starts with discovery of requirements. Then comes definition of the functional architecture and design of the core application. A small senior team of business analysts and system architects conducts these steps. A functional design is like a plan view for a building that shows the basic layout and infrastructure, but not all the details of each room. Plus, the functional design is already a working core application. From this core application, development builds in parallel, adding details. Teams are smaller than with conventional IT. Team roles encompass project leads, system architects, UI designers, knowledge modelers, software engineers, test and quality assurance, trainers, support personnel, and system administrators. Time to solution is two to ten times faster than with conventional IT development. Benefits start early. This reduces risk. Integration with existing systems and infrastructure is non-invasive. Development is iterative and incremental. Development can be highly parallel, but is significantly less labor-intensive. Testing and acceptance is ongoing rather than weighted towards the back end of the process. Deployment is incremental.
- 43. Concept computing lowers operating costs, total cost of ownership (TCO), and cost of maintenance compared to current operations. Based on customer experience reported by Be Informed, the rule of thumb is 30-60-90: Operating costs can be one-third less. Total cost of ownership can decrease by as much as two-thirds. Time and effort to make changes can decrease by up to 90 percent. Why? It’s much easier to integrate new data sources and interface new services by changing knowledge models than it is by writing code and rebuilding data stores.
- 44. Conclusion concept computing is coming! • paradigm shift. • Synthesis of AI, semantic, model-driven, mobile, and User interface technologies. • Architecture where Every aspect of a solution is semantic and model-driven • methodology Every stage of the lifecycle (development, operations, & evolution) is semantic and model-driven • new domain where value multiplies. For more information: www.project10x.com