Sources:
1. Big Data: “Gartner Press Release, "Gartner Reveals Top Predictions for IT Organizations and Users for 2012 and Beyond“ Dec 1, 2011
2. Analytics: “Predicts 2012 Information Infrastructure and Big Data,” Mike Blechar, Merv Adrian, Ted Friedman, W. Roy Schulte & Douglas Laney, 29 November 2011, #G00226066
3. Social Business: “Business Gets Social Innovation Key Initiative Overview”, Carol Rozwell, July 22 2011,#G00214406.
4. Smarter Planet: Pike Research, “Smart Cities: Intelligent Information and Communications Technology Infrastructure in the Government, Buildings, Transport, and Utility Domains”, 2Q11, p. 7
5. Mobile Enterprise: IBM New Workplace CIO Study, October 2011
6. Cloud: Information Week, Outlook 2012
7. Security: “Gartner Press Release, "Gartner Reveals Top Predictions for IT Organizations and Users for 2012 and Beyond“ Dec 1, 2011
8. Growth Markets: IDC “Asia/Pacific Excluding Japan 2012 Top ICT Predictions”, Dec 2011
Evolution of the “computer” from tool for calculation, to data warehouse, to delivery of information to knowledge
3 CHANGES to EMPHASIZE:
EMPHASIS ON DYNAMIC NATURE OF MODELS, NOT STATIC
- ACTIVE LEARNING – Hard
- DYNAMIC Engines (Training, Policy, Hypothesis, Outcome, Verification)
- Natural interfaces
How is our Learning System different from past Machine Learning approaches?
Our Learning System will automatically identify key features. Key Features selection is the technique of selecting a subset of relevant features for learning models. For example, key features to diagnose an illness may be a person's temperature, white blood cell count, pH level, etc. Current state of the art either has (A) humans identifying what are the key features for different domains or (B) allowing machine learning programs to extract key features based on expert rules (provided by humans) or statistical methods which may lead to false conclusions in domains that involve semantic ambiguity.
The Learning System we're building will use crowd sourcing techniques to automatically identify key features for a domain and will proactively ask humans for disambiguation, instead of waiting for humans to notice the model is erroneous (for example models that rated questionable mortgages as AAA or a software program that deduces that Internet Cookies are edible). In this vein, another key difference in our Learning System is active continuous verification. The current trends that provide increasing amounts of digital data (e.g. IBM Smarter Planet sensors) will enable our Learning System to modify itself to prune key features that are no longer relevant. In summary, (1) Automatic Extraction of Key Features (2) Continuous active self verification and (3) The ability to select the appropriate Machine Learning technique (statistical, genetic programming, neural networks, etc), and modify these techniques to changing conditions - all these three features have not been integrated into prior Machine Learning approaches.
A hypothesis is necessarily about a problem that is not formalized (if the problem were formalized, then no hypothesis would be required, only a formal solution)
Without a formal problem, the task of formulating hypotheses becomes one of creating alternative problem representations and selecting among them, in part, based on possible solutions to each
Known systems that attempt to do this require a defined problem space, where the range of possible hypotheses is calculated from a range of possible system states
“Real world” problems do not emerge from a range of possible states, however, but instead occur when previously defined ranges (or dimensions) are violated
The only known systems capable of formulating hypotheses about arbitrary states and selecting among them are biological cognitive systems
Explanation of this is necessary before a system that "Creates Hypotheses" can be introduced, even as a hypothetical
Some processes are inherently uncertain. We do not know with precision how long it takes to drive from A to B, nor how many and which chips on a wafer will pass quality tests.
Uncertainty comes from the data as well, in lots of ways: text (typos, ambiguities, conflict) – are these two people the same? Can I find evidence by looking at lots of data that they are, or are not? If I do then I have reduced that uncertainty. We have already talked some about sensors. It is well known that location-finding techniques do not work well in complex environments, like cities – we will talk more about that later, but the issue is a simple one. One would think that processing geospatial (GPS) data is core to managing a city – how good a job can I do if the location information is poor for assets like maintenance trucks, fire trucks, police.
Finally, there are new kinds of data uncertainty that come from things like social media: rumors, lies, falsehoods, wishful thinking. One has to distinguish the nuggets from the ore. One can look online to see many postings about contaminated baby food making its way from China to the US in 2008. Those rumors turned out not to be true – how does a system that processes petabytes of text understand these nuances – how to train it such a system.
Finally there are model uncertainties – we often approximate complex environments in order to be able to query them more efficiently – approximating collection of points with a line or forecasting a hurricane. But we must understand that using these models we do not have perfect answers, and we have to take into account these imperfections in business decisions. The good news is that we (IBM) have been managing a business with uncertain processes models, and data for many years – the semiconductor manufacturing business is based on driving wafer processes until the yield is good enough to scale into production – we are used to uncertainty bars here, at the manufacturing level, and at the micro level where we are studying the physical and chemical processes of nanostructures.
The challenge is to understand how we deal with uncertainty when we try to analyze big data – how do we represent uncertain data, reduce uncertainty of data, reason about that data in a way so that we can make business decisions (yes, no). In 2011 we demonstrated in a very public way a system that dealt with uncertainty and made business decisions – Watson playing Jeopardy. The system did a good job at understanding confidence in answers based on a variety of factors, in order to know whether it had the answer right or wrong. Of course sometimes Watson got it wrong. So the kinds of business decisions that we will make based on uncertain data need to be appropriate.
Can I route firetrucks if I do not know where they are (Data )what the road status is? How can I plan my plant capacity if my equipment is not predictably functional (Data,Model)? Do I know enough about a potential customer to be able to offer an appropriate sales incentive? [Data]. How about if I have 500,000 such customers? [Scale]
In another GTO topic, we talk about companies who's value is correlated with their ability to derive insight from their data: today we think about companies like Google or Facebook.
But we need to be thinking also about Healthcare companies (their medical records are potential business assets with huge value to them if they can monetize insight). Similarly, the ability of a retail company to be able to use its VIP loyalty data, its billing history, and maybe public social media to be able to do targeted marketing, or assess the acceptance of a product or a brand is potentially enormous.
When we look at where this explosive data growth is coming from we see that it is messy stuff: text, images, videos, audio, sensor data This is true inside the firewall as well. Manufacturers have large physical assets that are sensor-connected that have to be managed. Using these sensors with process models for equipment breakdown provides an ability to plan better. Better planning reduces costs.
On the next chart we will talk about some of these applications.
Create innovative capabilities that enable work to be performed anywhere, anytime, with anyone in a secure and socially collaborative way to further IBM's leadership as a smarter workforce
The time is ripe for Socially Synergistic Solutions, because of the increasing adoption of social software both in consumer space and by enterprises.
Fortune Global 100 Study:
Data was collected between November 2009 and January 2010 among the top 100 companies of Fortune痴 Global 500 companies. Sample size for countries/regions: U.S. = 29 companies, Europe = 48 companies, Asia-Pacific = 20 companies, Latin America = 3 companies. Because of the low sample size for Latin America, data is only broken out for this region for overall activity rates. Active accounts have at least one post in the past 3 months.Outliers have been noted. Data was collected by Burson-Marsteller global research team. (Burson-Marsteller (www.burson-marsteller.com), established in 1953, is a leading global public relations and communications firm. It provides clients with strategic thinking and program execution across a full range of public relations, public affairs, advertising and web-related services. The firm’s seamless worldwide network consists of 72 offices and 60 affiliate offices, together operating in 85 countries across six continents. Burson-Marsteller is a part of Young & Rubicam Brands, a subsidiary of WPP (NASDAQ: WPPGY).
We are at a tipping point in the convergence of social software and analytic technologies.
Businesses require a tight coupling of the two, in what we call “socially synergistic solutions.”
We are at a tipping point in the convergence of social software and analytic technologies, and business will increasingly require a tight coupling of the two to maximize their success. Customers will want to combine data from sensors and streams, their enterprise data, and social data -- that's data from or about people -- and use computation and analytics to bring the information together, aggregate, filter, and correlate it, and analyze it together to generate new insights. (as Pepsi, for example, wants to do to understand how their brands are perceived and how that influences sales in different demographics. And they will want to use analytics to more effectively take advantage of emerging techniques in social software to involve people, get their participation and improve their performance, and influence their behavior (as the city of Dubuque wants to do, by combining sensor data, analytics, and techniques taken from digital games, to reduce the use of resources like water.)These capabilities have traditionally existed in two different silos, with one set of technologies used to analyze a companies formal data like transactions, metrics, and KPIs, and another set of tools used to support the often tacit and unstructured human and collaborative work that goes on. And we've seen some blending of the two, initially with people emailing a spreadsheet, or setting up a discussion space to discuss data from analytic systems. And increasingly, systems like Cognos 10 and Manyeyes integrate collaboration tools with the reporting and visualization ones. What we are describing here goes beyond that, taking the deep analytic capabilities and applying them to the social data.