8/14/2019 US Army: poindexter-information-technologies-national http://slidepdf.com/reader/full/us-army-poindexter-information-technologies-national 1/20 Information, Technologies & National Security for the 21 st Century Neal Pollard Lecture Georgetown University 14 November 2007 Dr. John M. Poindexter [email protected]
20
Embed
US Army: poindexter-information-technologies-national
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
8/14/2019 US Army: poindexter-information-technologies-national
Data −> Information => Knowledge π> Options ε> Action
Where operators (functions) are:
−> Analysis = selects data in context to produce information
=> SenseMaking = understanding what the information means
π> PathFinding = deciding what to do about it in policy contextε> Execution = “operational forces” carry out decisions taken
Iteration = many steps are often repeated
• Simplified but basic non-linear process that is essential to understand.• Analysis is an over-used term.• This provides a working definition of Sensemaking and Pathfinding.
• Process carried out in a collaborative environment with relevant agencies.• Great deal of confusion amongst the terms data, information and knowledge.• “Operational Forces” – military, diplomatic, economic, law enforcement, etc.• This process is applicable in many other domains.
• Business, Financial, Medicine, Social Decisions to name a few.
Involves All of National Security Community Not Just Intelligence…
8/14/2019 US Army: poindexter-information-technologies-national
• Relational Data Bases – Primarily designed to answer static queries from structured data – in records and fields – Typical in the accounting, logistics, personnel areas – Tables are designed to answer the queries promptly – They take a big performance hit when fields relevant to a new query are in different tables – Table joining takes a lot of time – Unstructured text presents complications
• All words are not indexed – just those deemed significant by somebody in some general context
– Great way to store information, but not a good way to find relevant information in response to changing queries
• Traditional Search – Google is good example, but there are many others – When you submit a Google query it goes against a prepared linear index
– Lists of documents are returned – not entities (people, places, and things) – Again not all words in the documents are indexed – Difficult to focus in on relevant documents when context is complicated – Following one dimensional links in documents is a laborious process
• Situation – Enormous amounts of data being collected and stored in relational data bases
– But can’t find data of interest or it takes too long – Hard to find patterns – Dr. Tom Fingar, DDNI(Analysis), has said:
“It isn’t intelligence until it has been processed through the brain of an analyst. It’s
just data. And we are awash in data.
We don’t have enough analytical brains to meet all of the challenges. We have to
rely on technology.”
Traditional ways of finding information in data…
8/14/2019 US Army: poindexter-information-technologies-national
• Early mammals started sniffing around in the dark, so they neededa memory-map* based on olfaction to remember where things wereand predict where they might be.
• We have surely advanced beyond only our olfaction sensors, butthe process is instructional.
• Need memory-map of patterns (associations) and means for
exploring this map through: – Navigating (including precise fixes as well as dead-reckoning)
– Hunting (when looking for a specific target)
– Foraging (when targets are sparse and hard to find)
– Browsing (when high concentration of relevant data)• Scale of today’s problem is enormous.
– Most of scale is noise – uninteresting data
We should learn from our past about finding things…
*For macro theory on the way human brain works:On Intelligence, Jeff Hawkins with Sandra Blakeslee
8/14/2019 US Army: poindexter-information-technologies-national
• Humans remember and predict things through patterns of association.• Associative Memory problems in the past have been of scale & performance
– Must manage large N x N matrices for anything but toy problems. – Graffiti uses something on the order of N = 100 (Jeff Hawkins story).
– Can now manage on the order of N = 10 million. – And have on the order of 10 million entities with 107 x 107 memories. – Near instantaneous response to queries independent of memory base size
• Because N can now be so large, do not have to resort to dimensionality reductionfor complex problems. – Remember all associations from every observation.
– Do not throw away any information. – The more information the better. – However typical problem in this area is large number of entities with sparse data on
each buried in lots of noise which kills conventional approaches.
– Associative memory approach squeezes out every last ounce of information to provideintelligence.
• Have achieved very high ingestion rates.
• The memory base technology is a truly disruptive technology
Note: I’m a founding member of the Board of Directors of Saffron Technology,a producer of associative memory technology. www.saffrontech.com
Design influenced by the way we think the brain works…
8/14/2019 US Army: poindexter-information-technologies-national