Please see the attached document for details and a short summary.

Look forward to the presentation and follow up discussion.

Prof Robert L Fry on “Qualia, Intelligence, and Computation”

then turn away from the computer toward a blank wall or ceiling and stare at the same spot while blinking your eyes continuously

Robert L. Fry

Johns Hopkins Univeristy/Applied Physics Laboratory

entitled:

“Qualia, Intelligence, and Computation”

Thursday, 26 Jan, 1400 – 1450 hrs.

AFIT Building 640 Room 244

Abstract: (please also see attached document)

This talk is an update to another given to the AFIT Physics Department in
May 2007 titled “The Engineering of Intelligent Systems.” The emerging
framework described there has matured significantly since and is lends
itself to quantifying the notion of qualia and the active flow of
information and decisions to and from an open system. These systems include
intelligent, biological, dissipative physical systems, and “algorithms.”
Qualia can be seen as the most elementary computational tokens within this
general theory of computation and capture how a system mechanistically
acquires information from and in turn makes decisions on how it wants to
effect its environment. It should be noted that things that we can “do”
share the title of being “qualia” along with those things that we can
“know.” In either case, these are the percepts we subjectively distinguish.
Together, the requirements of subjective distinguishability and causality
provide a coherent and logically consistent basis for a computational
framework that describes the objective rules and dynamics of subjective
computation.
Progress since the prior talk includes answering two questions
critical to developing a complete engineering framework. The first was how
a system can solve the game-theoretic optimization problems so as to be
practically realizable. The second problem is drawn from so-called “rabbit
hole” analogy in turn drawn from the book “Alice in Wonderland” and deals
with how we should behave when confronted with uncertainty in what is known,
what to do, or both.
Once the engineering framework is summarized, examples and sample
applications are given in the areas of neural computation, ballistic missile
defense, algorithm design, and cancer research. Cortical neurons provide an
especially useful and constructive demonstration of the theory and
engineering framework. This talk ends with a synopsis of a recent
interesting finding. The term “qualia” and the fundamental notion and
definition of “strict” logical implication in logic have a common originator
- the great American philosopher C. I. Lewis. This may be surprising, but
should not since these concepts are inseparable. Lewis clearly understood
the importance of both notions.

About the speaker:

Robert Fry has bachelor degrees in computer science and electrical
engineering (EE) and a Masters degree in EE from the Johns Hopkins
University where he worked since 1979 and is currently Principal
Professional Staff. His principal job is combat and weapon system
engineering. He has extensive experience with the US Navy AEGIS combat
system and the development and evolution of many missile and weapon systems
including PATRIOT, THAAD, HARPOON, RAM, Tomahawk, AMRAAM, and the entire
Standard Missile family. Robert also performs basic research and is
transitioning a computational theory of intelligence into a formal framework
for the design and realization of intelligent systems. The transitioning
process has been guided by the two pathfinder applications; ballistic
missile defense (BMD) and neural computation where he has many publications
since 1994. In the latter, he has successfully applied the theory to
reverse-engineer cortical neurons as found in brains. Robert holds 5 US
patents including a Photoscreener for Infants and Preschool children now
being licensed and another in Cybernetic Systems. He teaches graduate
courses at Johns Hopkins University in Probability and Information Theory,
Random Processes, Digital Signal Processing, and various short courses in
missile and combat system engineering system taught domestically and
internationally.2012-01-26 Fry

This week we will start by answering any questions people have left over from the overview lectures from the last two weeks. So if you have concerns about that material jot them down and lets hash them out. The Second topic will be Memory. There have been several questions about what we mean by Working memory and long term memory. We will frame our discussion around three recent Scientific American articles.

1.) A Feeling for the Past – Emotion engraves the brain with vivid recollections but cleverly distorts your brain’s record of what really took place …So-called flashbulb memories actually fade considerably even though, paradoxically, they seem so vivid that we hold a misguided confidence in their fidelity … Putting a positive spin on a bad situation—a technique called cognitive reappraisal—can bothenhance accuracy in emotional memories and diminish their negative overtones …Happy memories are susceptible to distortion, too. We tend to recall fewer perceptual details of pleasant events, however, than of troublesome ones…what distinguishes flashbulb memories is “this sense of enhanced vividness and inflated confidence that we have in the accuracy, this sense that I will never forget ‘X.’”… For one thing, emotion is selective in how it enhances memory. Experts noticed long ago the “weapon focus effect”…Instead older adults appear to actively manage their emotions by paying less attention to negative things…How sleep meddles with memories, however, is complex…Another potential way to enhance accuracy in emotional memories while also damping down their negative overtones is to put a positive spin on a bad situation—a technique called cognitive reappraisal…
2.) Totaling Recall: manipulating / erasing memories – decades scientists believed that long-term memories were immutable—unstable for a few hours and then etched into the brain for good. Research now suggests that recalling a memory causes it to revert temporarily to an insecure state, in which the recollection can be added to, modified, even erased…possibility of deleting, or at least muting, parts of human memory with drugs or targeted therapies…To create, or consolidate, stable long-term memories, the brain must synthesize specific proteins in the hours after events occur…brief window before the memory is “reconsolidated,” it is susceptible to perturbation. “We used to think the memories we had were pictures of the original event. Now we know that it is the last version of the memory because each time we retrieve it, it changes a little bit,… Blocking the actions of the enzyme PKMzeta in rats wiped out the animals’ recollections of the event…
3.) Trying to Forget – Solomon Shereshevsky could recite entire speeches, word for word, after hearing them once. In minutes, he memorized complex math formulas, passages in foreign languages and tables consisting of 50 numbers or nonsense syllables. The traces of these sequences were so durably etched in his brain that he could reproduce them years later…weight of all the memories, piled up and overlapping in his brain, created crippling confusion. S could not fathom the meaning of a story, because the words got in the way…The problem with our memories is not that nothing comes to mind—but that irrelevant stuff comes to mind…attention-deficit hyperactivity disorder (ADHD) are also more likely to be among the forgetless (to coin a term). In short, memory—and forgetting—can shape your personality…reported that instructions to forget some learned items could enhance memory for others. Forgetting is therefore not a sign of an inferior intellect— but quite the opposite. The purpose of forgetting, he wrote, is to prevent thoughts no longer needed from interfering with the handling of current information…There’s a huge range in how effective people are at forgetting…

This meeting will consist of Capt Amerika giving the QUEST overview talk in its entirety. For those who have never heard the whole story this is a great opportunity to get caught up, and for those who have heard it before we look forward to any and all criticisms/suggestions for how to improve upon it.

Slides to come for those interested, let me know and I’ll distribute them individually.

QUEST Point Paper (Qualia Exploitation of Sensing Technology)
Jan 6, 2012
Dr. Steve Rogers (‘Capt Amerika’)

Walk Away Points – QUEST – Cognitive Exoskeleton for Intelligence Amplification, IA not AI:
1.) Purpose of QUEST: Need for solutions to facilitate decision making in many areas – ISR/ PCPAD (layered sensing = ‘what to look at’), Cyber, ISHM, Cognitive EW, SSA, Medical, … . In all these areas we don’t lack data – in fact they all suffer from ‘drowning in data’. We lack relevant information! We don’t expect a magic bullet to replace the human thus the need for an integrated human/computer solution = flexible autonomy.

2.) What is unique about QUEST: traditional approaches put the computer between the world (cloud of potentially useful sensors and data) and the human. This results in – What is it doing? Why is it doing that? What will it do next? Quest solutions designed to allow the human / computer to align with each other, wingman solutions. The computer has a representation of the human’s representation of the battle space and vice versa, Theory of Mind. Along these lines we have an agent specific definition of data, information and context.

3.) Key characteristics of QUEST solutions:

a. Dual Process theory – intuition system (sys1) blended with deliberation system (sys2).

i. Autistic Sys1 (Intuition – Data Driven Artificial Intelligence, DDAI) – facilitate quick reflexive responses to stimuli that are close to prior experiences, signature based recognition. Conclusions of decisions made in sys1, gut or hunch decisions, are posted to the Sys2 representation without the details of the methodology, intuition.

ii. Deliberation Sys2 (‘feeling’ = Qualia) – multi-modal subjective simulation based narratives (artificial conscious representation where fundamental unit of cognition is a situation used for perception, recollection and for imagination) to allow handling the unexpected Query (includes a pattern completion inference mechanism – most of the concepts in perception are inferred not captured by sensors) and the key to re-programming of Sys1. Situated Conceptualization – Qualia theory of Relativity = red doesn’t come from the object it is created in your mind, making a computer that can ‘feel’. The conceptualizer is placed in the representation and provides the ‘feeling’ of self.

b. Theory – a common mathematical framework for both the human and the computer decision aid allows for theoretical bounding of performance of the integrated solution – a Theory of Knowledge (what can this team of humans/computers know?).

c. Close the Loop – interaction with the environment is the key to the generation of a stable, consistent and useful representation. Plausible narratives compete for attention and are continually refined based on their stability, consistency and usefulness.

Application Details ‘where are we now?’

4.) ISR/PCPAD – recently initiated -WAMI/FMV where do I look? How do I process all this extras data without increasing the number of human analysts? (Mission Driven Exploitation) – solve the layered sensing / D2D challenge – PCPADx cognitive unification, analyst pull

5.) Cyber – We have young airmen in the Network Defense business watching screens not unlike those in the DGSs watching video. We have malware detection approaches that are autistic. QUEST ‘unknown malware’ approaches allow humans to make better decisions because they can align the information provided to them to the decisions they have to make.

6.) Integrated Systems Health Monitoring – We have F-15s falling out of the sky and modern composite aircraft that we don’t know how to maintain. Addressing these issues by permeating the structures with sensors leads us back to the data overload issue and don’t address the fundamental question, ‘Can I do this mission in this aircraft at this moment?’ Initiated effort to bench test quest solutions.

Walk-away:
Application domain agnostic flexible autonomy solutions to Sensor Data Exploitation are the key for a revolutionary breakthrough in the drowning in data problem.

Currently 125 people are active on the weekly mailings from across the country! Mixture of academics, industry and government scientists and engineers.

***seems to imply that they can no longer ‘situate’ the features – thus a loss of the relevant aspects of sys2 associated with visual information processing for these types of ‘blobs’ ***
Face deficit holds object lesson
Recognizing mugs may not be special in the brain
By Bruce Bower
Web edition : Friday, December 16th, 2011

Enlarge
OBJECTS OF CONFUSION
Objects of Confusion A brain-damaged man who can’t recognize faces also fared poorly at learning to distinguish computer-generated “Greebles” (such as these) from one another. Each Greeble displays a distinctive configuration of three appendage types.C. Bukach
A brain-damaged man who can’t remember faces has nosed into a scientific debate about how people learn to recognize other complex objects. Deaf users of sign language also have a hand in this dispute.

The brain-damaged man’s facial failures are one symptom of a general inability to perceive configurations of object parts, suggests a new investigation led by psychologist Cindy Bukach of the University of Richmond in Virginia. The man thus stumbles at identifying not only people’s faces but also computer-generated, three-part objects called Greebles, even after extensive training, Bukach’s team reports online December 8 in Neuropsychologia.

Bukach and her colleagues studied LR, a man who fails to recognize his daughter when shown a picture of her but remembers distinctive facial features, such as Elvis’ sideburns. Damage in a car accident to a brain area just under the right temple caused this condition, called prosopagnosia.

“There are many ways in which face recognition can be disrupted, but our evidence shows that LR’s type of prosopagnosia impairs recognition of objects with multiple parts, with faces as the most obvious example,” Bukach says. Relative positions of the eyes, nose and mouth, as well their shapes, contribute to perceiving a face as a single entity.

In a 2006 report, her team designed a collection of eight faces using different combinations of two sets of eyes, noses and mouths. After briefly viewing a face, LR correctly selected it from all eight faces 25 percent of the time — about what would be expected if he based choices on a single facial feature, Bukach says. Further testing showed that LR homed in on the mouth.

In the new study, the researchers designed eight Greebles, using different combinations of two versions of three distinctive appendages. LR recognized Greebles he had just seen 31 percent of the time, improving little after several one-hour, weekly training sessions. Four healthy volunteers struggled at discerning Greebles at first but recognized most of them after training.

Bukach opposes an influential view that the brain evolved systems for dealing with key types of knowledge, including face recognition (SN: 7/7/01, p. 10). A proponent of that view, psychologist Bradley Duchaine of Dartmouth College, previously reported that a prosopagnosia patient named Edward — who cited lifelong problems recognizing faces — learned to discriminate Greebles but not human faces.

If face recognition depends on a general capacity for learning to recognize multi-part objects, Duchaine holds, healthy volunteers should recognize novel Greebles as poorly as prosopagnosia patients do at first but perform better than patients after seeing lots of Greebles. LR’s Greeble difficulties exceeded those of healthy volunteers from the start, a sign of fundamental object-recognition problems that make the results hard to interpret, Duchaine contends. “These new results don’t help us understand mechanisms used for face processing,” he says.

LR’s poor Greeble-detection accuracy before and after training indicates that he focused on only one Greeble appendage when trying to tell the funny-looking objects apart, Bukach responds.

Support for the idea that brains use a general mechanism to recognize complex objects comes from deaf people who communicate with American Sign Language. Just as upside-down faces look weird and often unrecognizable to healthy volunteers, so do upside-down signs shown to fluent ASL users, say psychologists David Corina of the University of California, Davis, and Michael Grosvald of the University of California, Irvine.

Because healthy individuals perceive faces as whole entities, topsy-turvy faces look bizarre, Corina says. Likewise, ASL users learn to see signs as integrated sets of movements that look peculiar when inverted, the researchers propose in a paper published online December 6 in Cognition.

Many researchers assume that people understand sign language by breaking each sign down into hand shapes, arm movements and other elements.

Corina and Grosvald also find that deaf ASL users are faster than hearing nonsigners at recognizing videos of head scratching and other common grooming actions. Sign languages exploit brain areas devoted to detecting human actions in general, they propose.

Psycholinguist Karen Emmorey of San Diego State University calls new evidence that fluent signers perceive signs as whole entities “a key insight.” Further work needs to confirm that learning a sign language modifies action-related brain areas, she adds

Inaugural Cognitive Brown Bag Lunch Gathering

When: Tuesday, December 20, 2011 11:30 AM-1:00 PM (UTC-05:00) Eastern Time (US & Canada).

Where: TecEdge, 500 Springfield Street, Suite 100, Dayton, Ohio 45431

Note: The GMT offset above does not reflect daylight saving time adjustments.

*~*~*~*~*~*~*~*~*~*

You are cordially invited to attend the inaugural gathering of the new RH Cognitive Modeling Brown Bag lunch series.

Date: Tuesday, 20 December 2011

Time: 1130-1300

Location: TecEdge, 500 Springfield Street, Suite 100, Dayton, Ohio 45431

As Kevin Gluck described in an earlier email, the purpose of this Brown Bag series is to foster a strong community and an atmosphere of collaboration in support of the growing research in cognitive modeling. To that end, the topic for this first gathering will be, primarily, a meet-and-great between people across the Human Effectiveness Directorate as well as some researchers outside of RH who pursue related research. Kevin Gluck (RHAC) will help provide us with an overview of some goals of the cognitive modeling expansion effort, and we’ll discuss the goals and future plans for the Brown Bag series itself.

Please do plan to bring your own lunch. TecEdge has drinks available (soda, water, coffee, tea) for 50 cents an item. Also, please bring any ideas or wants for future Brown Bag gatherings; your input will help us make these gatherings as effective as possible for our research community.

This gathering is open to any AFRL researchers with an interest in cognitive modeling, so if you know anyone who might be interested in attending, please feel free to forward this invitation along.

Please note, if you are not interested in being invited to future Brown Bag gatherings, please let me know and I will remove your name from the distribution list.

If you have any questions,

I look forward to meeting you all and kicking off a lively Cognitive Modeling Brown Bag series.

Cheers,

Leslie

Leslie M. Blaha, Ph.D

Engingeering Research Psychologist

711 HPW/RHCV

937-255-0425 Office

785-0425 DSN

Leslie.Blaha@wpafb.af.mil