Date post: | 18-Dec-2015 |
Category: |
Documents |
Upload: | adele-watkins |
View: | 212 times |
Download: | 0 times |
AFRL RAIR Lab Kickoff 6.10.04
Selmer BringsjordKonstantine Arkoudas, Yingrui Yang, Marc Destefano,
Paul Bello, Andy Shilladay, Josh Taylor, Bettina Schimanski
Rensselaer AI & Reasoning (RAIR) Laboratory
Department of Cognitive Science
Department of Computer Science
Department of Decision Sciences & Engineering Systems
Rensselaer Polytechnic Institute (RPI)
Troy NY 12180 USA
6.10.04
PART I: RAIR LAB OVERVIEW/TOUR
10a-1230p
The Rensselaer AI & Reasoning Lab(The RAIR Lab)
A while back,RPI StrategicInvestment
Cracking Project; “Superteaching”
Slate (IntelligenceAnalysis)
Item generation
synthetic characters/psychological time
Wargaming
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
hypothesis generation; AI in support of IA
Engineering Method for RAIR Lab’sNext-Generation Logic-based AI
• Isolate and dissect human ingenuity.(psychology of reasoning)
• Mathematize a weak correlate to this ingenuity courtesy of advanced logical systems.• Implement this correlate in working programs.• Augment the correlate with machine-specific power.
Engineering Method for RAIR Lab’sNext-Generation Logic-based AI
• Isolate and dissect human ingenuity.(psychology of reasoning)
• Mathematize a weak correlate to this ingenuity courtesy of advanced logical systems.• Implement this correlate in working programs.• Augment the correlate with machine-specific power.
the thinking enemy
our wargamers
...
But RAIR L overview first...
Next-Generation Logic-based AI
Reasoning Software:Systems to augmentand (sometimes) matchhuman reasoning-basedactivity.
Robot Reasoning:Robots able to accomplish impressive things on the strength of reasoning.
The Foundations ofAI & CogSci:Are people computers?Does Gödel’s incompletenessresults imply that minds aresuperior to all machines?...
PERI
Robot Reasoning R&D:PERI
(Psychometric Experimental Robotic Intelligence)
• Scorbot-ER IX
• Sony B&W XC55 Video Camera
• Cognex MVS-8100M Frame Grabber
• Dragon Naturally Speaking Software
• NL (CARMEL & RealPro?)
• BH8-260 BarrettHand Dexterous 3-Finger Grasper System
Robot Reasoning R&D:
PERI
QuickTime™ and aDV/DVCPRO - NTSC decompressor
are needed to see this picture.
Next-Generation Logic-based AI
Reasoning Software:Systems to augmentand (sometimes) matchhuman reasoning-basedactivity.
Robot Reasoning:Robots able to accomplish impressive things on the strength of reasoning.
The Foundations ofAI & CogSci:Are people computers?Does Gödel’s incompletenessresults imply that minds aresuperior to all machines?...
PERI ATP-Powered Bots
Next-Generation Logic-based AI
Reasoning Software:Systems to augmentand (sometimes) matchhuman reasoning-basedactivity.
Robot Reasoning:Robots able to accomplish impressive things on the strength of reasoning.
The Foundations ofAI & CogSci:Are people computers?Is it possible to formally modelthe ethical and epistemic attitudesof human beings? What aboutevil -- can it be mathematized?...
Next-Generation Logic-based AIReasoning Software:Systems to augmentand (sometimes) matchhuman reasoning-basedactivity.
Robot Reasoning:Robots able to accomplish impressive things on the strength of reasoning.
The Foundations ofAI & CogSci:Are people computers?Is it possible to formally modelthe ethical and epistemic attitudesof human beings? What aboutevil -- can it be mathematized?...
Next-Generation Logic-based AI
Reasoning Software:Systems to augmentand (sometimes) matchhuman reasoning-basedactivity.
Robot Reasoning:Robots able to accomplish impressive things on the strength of reasoning.
The Foundations ofAI & CogSci:Are people computers?Does Gödel’s incompletenessresults imply that minds aresuperior to all machines?...
e.g., Slate
The Slate System (v1.4)
Next-Generation Logic-based AI
Reasoning Software:Systems to augmentand (sometimes) matchhuman reasoning-basedactivity.
Robot Reasoning:Robots able to accomplish impressive things on the strength of reasoning.
The Foundations ofAI & CogSci:Are people computers?Does Gödel’s incompletenessresults imply that minds aresuperior to all machines?...
e.g., Slate
GameDevelopment
Game Development in the RAIR Lab
&
Teamed Up w/ VV (and, for ARDA, Planet 9)
Next-Generation Logic-based AI
Reasoning Software:Systems to augmentand (sometimes) matchhuman reasoning-basedactivity.
Robot Reasoning:Robots able to accomplish impressive things on the strength of reasoning.
The Foundations ofAI & CogSci:Are people computers?Does Gödel’s incompletenessresults imply that minds aresuperior to all machines?...
e.g., Slatesoftware forwargaming
GameDevelopment
The RAIR Lab OffersSix Interconnected Benefits for
Wargaming & Military Simulation:• The most sophisticated machine reasoners: Athena, MARMML/Chogic,
(and “souped up” classics like SNARK, Otter, OSCAR, Vampire)– six attributes tailor-made for the demands of advanced wargaming (handles
beliefs, knowledge, ethics, temporal operators, etc.)
• Symbiotic tie-in, in any R&D conducted for and with AFRL, the ARDA-sponsored Slate system
• Command over commercial games, including wargames/strategy games, etc.
• The capacity to build advanced synthetic characters for wargames– on the basis of these machine reasoners, and, for the “easy” processing, ACT-R,
Soar
• The capacity to engineer transparent systems, including transparent virtual environments in which the effects of actions can be completely charted and understood
• A concrete marriage of the math behind decision-making with the math behind reasoning
PART II: OVERVIEW OF LOGIC-BASED AI
1245p-145p
Nilsson’s (Simple) Overview
Knowledge-Based Agents(AIMA/AIMA2e)
J-L 1
Suppose that the following premise is true:
If there is a king in the hand, then there is an acein the hand, or else if there isn’t a king in the hand,then there is an ace.
What can you infer from this premise?
There is an ace in the hand.NO! NO!
In fact, what you can infer is that there isn’t an ace in the hand!
Cracked Easily in Natural Deduction
Brief Interlude on the Propositional Calculus &
First-Order Logic...
Scenarios forIntelligence Analysis
Wargaming,Simulated C2,Military Simulations
“New Order” Microscenario #1(“no distractor” version)
John H. was killed by a member of the Al-Qaeda cell 'The New Order'.
The only members of 'The New Order' were John H., Majed H., and Essid D.
Within-cell killings only occur when the attacker believes the victim is a traitor, and never when the attacker is of lower rank.
Essid D. believes that nobody is a traitor who John H. believes is a traitor.
John H. believes everyone except Majed H. is a traitor.
Majed H. believes that everyone who is not of lower rank than John H. is a traitor.
Majed H. believes everyone is a traitor who John H. believes is a traitor.
No one believes everyone in 'The New Order' is a traitor.
‘John H.’ is not an alias for ‘Majed H.’, nor vice versa. In addition, ‘Majed’isn’t an alias for ‘Essid’ (nor, again, vice versa).
Subject Tacking “New Order #1”
QuickTime™ and aVideo decompressor
are needed to see this picture.
“New Order” Microscenario #1(“no distractor” version)
John H. was killed by a member of the Al-Qaeda cell 'The New Order'.
The only members of 'The New Order' were John H., Majed H., and Essid D.
Within-cell killings only occur when the attacker believes the victim is a traitor, and never when the attacker is of lower rank.
Essid D. believes that nobody is a traitor who John H. believes is a traitor.
John H. believes everyone except Majed H. is a traitor.
Majed H. believes that everyone who is not of lower rank than John H. is a traitor.
Majed H. believes everyone is a traitor who John H. believes is a traitor.
No one believes everyone in 'The New Order' is a traitor.
‘John H.’ is not an alias for ‘Majed H.’, nor vice versa. In addition, ‘Majed’isn’t an alias for ‘Essid’ (nor, again, vice versa).
Solved By Hand in Hyperproof
Slate Used to Crack “New Order #1”
Using Athena to: Find out who killed; automatically obtain a proof; construct and check a natural deduction-style proof
(define culprit-property (forall ?x (iff (culprit ?x)
(killed ?x John))))
(assert culprit-property)
(find-model (add (exists ?x (culprit ?x)) (ab)))
(!prove (killed John John))
((killed John John) BY (!by-contradiction (assume (not (killed John John)) (dlet ((disjunction (!derive (or (killed Essid John) (killed Majed John)) [(not (killed John John)) premise1 premise2]))) (!by-cases (assume (killed Essid John) (dlet ((S1 (!derive (believesTraitor Essid John) [premise3 premise2 (killed Essid John)])) (S2 (!derive (believesTraitor John John) [premise5 premise9])) (S3 (!derive (not (believesTraitor Essid John)) [S2 premise4]))) (!derive false [S1 S3]))) (assume (killed Majed John) (dlet ((S1 (!derive (believesTraitor Majed John) [premise3 premise2 (killed Majed John)])) (S2 (!derive (believesTraitor John John) [premise5 premise9])) (S3 (!derive (believesTraitor John Essid) [premise5 premise9])) (S4 (!derive (believesTraitor Majed Essid) [S3 premise7])) (S5 (!derive (not (believesTraitor Majed Majed)) [S1 S4 premise8 premise2])) (S6 (!derive (not (lowerRank Majed John)) [(killed Majed John) premise2 premise3])) (S7 (!derive (believesTraitor Majed Majed) [S6 premise6]))) (!derive false [S5 S7]))) [disjunction])))))
Denotational Proof Languages (DPLs)
• DPLs are languages for writing proofs and proof tactics in arbitrary logics
• Novel syntax and semantics (based on the abstraction on assumption bases) ensure:– Readability and writability
– Efficient proof checking
– Guaranteed soundness
– Powerful mechanisms for expressing complex proof tactics and tacticals
Wide applicability
• DPLs have been designed and implemented for:– Classical logics (both first- and higher-order)– Intuitionist logics– Modal and temporal logics– Program logics (Hoare-Floyd logics)– Type systems
Athena
• A DPL for classical first-order logic
• Uses natural deduction
• Incorporates a higher-order functional programming language with algebraic data types
• Supports induction, recursion, pattern matching
• Other logics (e.g. modal logic) can be rapidly prototyped by implementing them on top of Athena
PART III: WARGAMING AND ADVANCED SYNTHETIC
CHARACTERS145p-230p
PART IV: WARGAMING AND ADVANCED SYNTHETIC
CHARACTERS230p-330p
Wargaming Formalized?
• Don’t yet have a formal account.– lots of books and papers... but not a lot of rigor
• But -- we know that agents are required.• We want agents that have human-level thinking
power:– We want advanced synthetic characters for wargaming
and military simulations– We want to model the mindset of terrorists, replete with
their ethical norms, vs. ours, and replete with what they believe about us, what they believe about what we believe, what we believe about what they believe about what we believe, and so on
Building a Taxonomy of Wargames
Advanced Synthetic Characters: Background Reading
A classic originally published in 1946. Egri shows that at the core of all good dramatic writing (whatever its form) stand not rules for cranking out text, but fully developed characters. This book introduces the so-called "dialectical method," and connects it to case studies created on the fly, and to great drama of the past (e.g., Ibsen, and e.g. his immortal Nora). From the standpoint of AI and the creation of advanced synthetic characters, the book is daunting, as it asserts that to be a decent playwright one must have a monstrous amount of knowledge drawn from psychology, sociology, economics, and so on.
Kress breaks down the complex art of writing into numerous techniques of representation. The first third of the book concentrates on techniques of characterization. From an AI stance, the book presents a few interesting challenges: it presents evidence that a character’s exterior presentation must be tightly bound to his or her history, and it asserts that all truly developed characters must be based on the author’s own internal emotional state and life experiences.
An adaptation of Stanislavsky’s Method for actors to writing, Collins focuses on the presentation of characters within narrative. The techniques it includes for demonstrating emotion through action appear readily applicable to the representational aspects of ASCs. However, much of the information about specific emotion is assumed to be drawn from the author’s personal life experiences, making some of its techniques difficult to apply.
Halperin’s eight chapters can be considered as separate essays, each tackling one aspect of characterization. Of particular use in ASCs are the chapters on interior motivation and cultural legacy, which provide useful “template” information to set a character within an internal and social context.
A dictionary of character details, McCutcheon is of primary use in populating knowledge bases for use in character generation. It provides a useful reference of physical traits, mannerisms, modes of dress and common names from which the groundwork for deeper representation could be laid.
Edelstein is a “cookbook” of character traits, organized by types of characters they are appropriate to. Within the confines of AI research, it useful in the sub-categorization of broad character traits into more specific associated details.
Hood provides a sequence of 3-4 page treatments of specific emotions, focusing on how to convey them effectively in prose. Although the book is primarily concerned with the language used to represent them, its discussion of emotional impact on behavior makes it useful in the generation of ASCs.
Primarily a guide to script-writing, Wolff contains a single chapter on creating three-dimensional characters that provides a first-draft structure for representing knowledge about a character. The current under-construction vMEM ASC is based on the 31-question overview of a character provided here as a starting point on which to base a Q/A system.
A collection of short essays on writing, Dickson contains a great deal of shallow and stereotypical information. Many of its essays on characterization are better considered as tactics to avoid using – they seem to favor quick solutions over deep representation. However, its discussion of the importance of central traits and character flaws in creating empathy is significant for the deep representation of such traits in ASCs.
Adv. Synthetic Characters: Background Read. Con.
Synthetic Characters To Leapfrog?
Same Thing Here:Definitely Not an Advanced SC!
Every behavior that happens inThe Sims is computed from a number(1- 10) for each attribute.
Where’s the cognition?
RASCALS ecumenical
RASCALS logic-based
First-Order Logic
Subsumption-Based Architecture
First Steps(w/ contract, SOW, $)
• Model “deontically intense” situation in game; implement; demonstrate (for our sponsors); refine; model...
• Model “epistemically intense” situations in logicist fashion; implement; demonstrate (for our sponsors); refine; model...
THE END
Decision-making meets Reasoning ...
“MARMML and Newcomb’s Problem”(separate ppt and papers)
Slate Hypothesis Generation in our Narrative Scenario ( v)
What is the destination of the convoy?
customary destinations ruled out
---------------- PROOF ----------------1 [] -Yar(x)|Terrorists(x).2 [] -WindAccessible(x,y)| -USBase(x)| -Bioagents(z)| -Terrorists(z)|AttackPosition(y,z,x).3 [] -CaveSystem(x,aconvoy)| -Accessible(x,aconvoylocation).4 [] -Camp(x,aconvoy)| -Accessible(x,aconvoylocation).5 [] -Village(x,aconvoy)| -Accessible(x,aconvoylocation).6 [] -AttackPosition(x,y,z)| -Convoy(y)| -Terrorists(y)| -PresentLocation(y,u)| -Accessible(x,u)|CaveSystem($f1(x,y,u,z),y)|Village(z2,y)|Camp(z3,y)|Destination(x,y).7 [] -AttackPosition(x,y,z)| -Convoy(y)| -Terrorists(y)| -PresentLocation(y,u)| -Accessible(x,u)|CaveSystem($f1(x,y,u,z),y)|Village(z2,y)|Accessible(z3,u)|Destination(x,y).8 [] -AttackPosition(x,y,z)| -Convoy(y)| -Terrorists(y)| -PresentLocation(y,u)| -Accessible(x,u)|CaveSystem($f1(x,y,u,z),y)|Accessible(z2,u)|Camp(z3,y)|Destination(x,y).9 [] -AttackPosition(x,y,z)| -Convoy(y)| -Terrorists(y)| -PresentLocation(y,u)| -Accessible(x,u)|CaveSystem($f1(x,y,u,z),y)|Accessible(z2,u)|Accessible(z3,u)|Destination(x,y).10 [] -AttackPosition(x,y,z)| -Convoy(y)| -Terrorists(y)| -PresentLocation(y,u)| -Accessible(x,u)|Accessible($f1(x,y,u,z),u)|Village(z2,y)|Camp(z3,y)|Destination(x,y).11 [] -AttackPosition(x,y,z)| -Convoy(y)| -Terrorists(y)| -PresentLocation(y,u)| -Accessible(x,u)|Accessible($f1(x,y,u,z),u)|Village(z2,y)|Accessible(z3,u)|Destination(x,y).12 [] -AttackPosition(x,y,z)| -Convoy(y)| -Terrorists(y)| -PresentLocation(y,u)| -Accessible(x,u)|Accessible($f1(x,y,u,z),u)|Accessible(z2,u)|Camp(z3,y)|Destination(x,y).13 [] -AttackPosition(x,y,z)| -Convoy(y)| -Terrorists(y)| -PresentLocation(y,u)| -Accessible(x,u)|Accessible($f1(x,y,u,z),u)|Accessible(z2,u)|Accessible(z3,u)|Destination(x,y).14 [] -Destination(amountain46,aconvoy).15 [] Convoy(aconvoy).16 [] Yar(aconvoy).17 [] PresentLocation(aconvoy,aconvoylocation).19 [] Accessible(amountain46,aconvoylocation).20 [] WindAccessible(amilbase33,amountain46).21 [] Bioagents(aconvoy).22 [] USBase(amilbase33).23 [hyper,16,1] Terrorists(aconvoy).24 [hyper,20,2,22,21,23] AttackPosition(amountain46,aconvoy,amilbase33).25 [hyper,24,13,15,23,17,19,unit_del,14] Accessible($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoylocation)|Accessible(z2,aconvoylocation)|Accessible(z3,aconvoylocation).26 [hyper,24,12,15,23,17,19,unit_del,14] Accessible($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoylocation)|Accessible(z2,aconvoylocation)|Camp(z3,aconvoy).27 [hyper,24,11,15,23,17,19,unit_del,14] Accessible($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoylocation)|Village(z2,aconvoy)|Accessible(z3,aconvoylocation).28 [hyper,24,10,15,23,17,19,unit_del,14] Accessible($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoylocation)|Village(z2,aconvoy)|Camp(z3,aconvoy).29 [hyper,24,9,15,23,17,19,unit_del,14] CaveSystem($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoy)|Accessible(z2,aconvoylocation)|Accessible(z3,aconvoylocation).30 [hyper,24,8,15,23,17,19,unit_del,14] CaveSystem($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoy)|Accessible(z2,aconvoylocation)|Camp(z3,aconvoy).31 [hyper,24,7,15,23,17,19,unit_del,14] CaveSystem($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoy)|Village(z2,aconvoy)|Accessible(z3,aconvoylocation).32 [hyper,24,6,15,23,17,19,unit_del,14] CaveSystem($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoy)|Village(z2,aconvoy)|Camp(z3,aconvoy).33 [hyper,26,4,25,factor_simp,factor_simp] Accessible($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoylocation)|Accessible(z2,aconvoylocation).34 [hyper,27,5,33,factor_simp] Accessible($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoylocation)|Accessible(z3,aconvoylocation).35 [hyper,28,5,33,factor_simp] Accessible($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoylocation)|Camp(z3,aconvoy).36 [hyper,35,4,34,factor_simp] Accessible($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoylocation).37 [hyper,29,3,36] Accessible(z2,aconvoylocation)|Accessible(z3,aconvoylocation).38 [hyper,30,3,36] Accessible(z2,aconvoylocation)|Camp(z3,aconvoy).39 [hyper,38,4,37,factor_simp] Accessible(z2,aconvoylocation).40 [hyper,31,5,39] CaveSystem($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoy)|Accessible(z3,aconvoylocation).41 [hyper,40,3,36] Accessible(z3,aconvoylocation).42 [hyper,32,5,39] CaveSystem($f1(amountain46,aconvoy,aconvoylocation,amilbase33),aconvoy)|Camp(z3,aconvoy).43 [hyper,42,3,36] Camp(z3,aconvoy).44 [hyper,43,4,41] $F.------------ end of proof -------------
Shows that mountain46 is convoy’s destination
Resolution Athena English
• Automatic generation of proofs in natural language, roughly in the same style that one encounters in rigorous proofs appearing in mathematical texts.
• Coupled with the automatic generation of counter-examples (in the form of finite models), such a feature should greatly help engineers building digital systems.
• Automatically generated counter-examples will help to catch bugs in the early stages of design and implementation; automatically generated proofs expressed in English will validate their design and implementation choices in later stages by demonstrating why the systems work.
Athena has just proved the UNIX OS sound! A lightning-fast
6000-long proof.
Simple Reasoning Problem
Everyone loves anyone who loves someone.
Alvin loves Bill.
Can you infer that everyone loves Bill?
ANSWER:
JUSTIFICATION:
(assert '(alive marc) :name 'marc-alive)(assert '(birthtime marc (date-point 1977 2 10 9 24)) :name 'marc-birthtime)(assert '(biological-mother regina marc) :name 'marc-mother)(assert '(biological-father josephjr marc) :name 'marc-father)(assert '(sister christine marc) :name 'marc-sister)(declare-predicate-symbol 'parent 2 :falsify-code 'irreflexivity-falsifier)(assert '(forall (?person) (not (parent ?person ?person))) :name 'parent-irreflexive)(assert '(forall (?person) (iff (parent ?person) (exists (?person1) (parent ?person ?person1)))) :name 'parent-unary-defintion)(assert '(forall (?person1 ?person2) (iff (parent ?person1 ?person2) (child ?person2 ?person1)))
:name 'parent-child-inverse)(assert '(forall (?person1 ?person2) (iff (parent ?person1 ?person2) (or (biological-parent ?person1 ?person2) (adoptive-parent ?person1 ?person2) (step-parent ?person1 ?person2) (foster-parent ?person1 ?person2)))) :name 'parent-subdivision)(assert '(forall (?person) (iff (mother ?person) (exists (?person1) (mother ?person ?person1)))) :name 'mother-unary-defintion)(assert '(forall (?person1 ?person2) (iff (mother ?person1 ?person2) (and (parent ?person1 ?person2) (female ?person1)))) :name 'mother-binary-defintion))(assert '(forall (?person) (exists (?time-interval) (lifespan ?person ?time-interval))) :name 'all-persons-have-lifespan)(assert '(forall (?person ?time-point ?time-interval) (iff (alive-at-time ?person ?time-point) (and (lifespan ?person ?time-interval) (temporally-intersects ?time-interval ?time-point)))) :name 'define-alive-at-time-point)(assert '(forall (?person) (iff (alive ?person) (alive-at-time ?person now))) :name 'define-alive)
vMEMInitially, a Q/A theorem proving-based system in which Questions will be answered by deducing Answers fromthe knowledge basecorresponding to vMEM. This knowledge base will be constructed in keeping with the construction of “deep”
characters in narrative. QuickTime™ and a decompressorare needed to see this picture.