unifying logic and probability a new dawn for artificial intelligence?
DESCRIPTION
Unifying logic and probability A new dawn for artificial intelligence?. AI : intelligent systems in the real world. AI : intelligent systems in the real world. The world has things in it!!. AI : intelligent systems in the real world. The world has things in it!!. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/1.jpg)
1
Unifying logic and probability A new dawn for artificial intelligence?
Stuart RussellComputer Science Division, UC Berkeley
LIP6, Université Pierre et Marie Curie
![Page 2: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/2.jpg)
2
AI: intelligent systems in the real world
![Page 3: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/3.jpg)
3
AI: intelligent systems in the real world
The world has things
in it!!
![Page 4: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/4.jpg)
4
AI: intelligent systems in the real world
Good Old-Fashioned AI: first-order logic
The world has things
in it!!
![Page 5: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/5.jpg)
Why did AI choose first-order logic? Provides a declarative substrate
Learn facts, rules from observation and communication Combine and reuse in arbitrary ways
5
![Page 6: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/6.jpg)
Why did AI choose first-order logic? Provides a declarative substrate
Learn facts, rules from observation and communication Combine and reuse in arbitrary ways
Expressive enough for general-purpose intelligence It provides concise models, essential for learning
6
![Page 7: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/7.jpg)
Why did AI choose first-order logic? Provides a declarative substrate
Learn facts, rules from observation and communication Combine and reuse in arbitrary ways
Expressive enough for general-purpose intelligence It provides concise models, essential for learning E.g., rules of chess (32 pieces, 64 squares, ~100 moves)
7
![Page 8: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/8.jpg)
Why did AI choose first-order logic? Provides a declarative substrate
Learn facts, rules from observation and communication Combine and reuse in arbitrary ways
Expressive enough for general-purpose intelligence It provides concise models, essential for learning E.g., rules of chess (32 pieces, 64 squares, ~100 moves)
~100 000 000 000 000 000 000 000 000 000 000 000 000 pages as a state-to-state transition matrix (cf HMMs, automata)
R.B.KB.RPPP..PPP..N..N…..PP….q.pp..Q..n..n..ppp..pppr.b.kb.r
8
![Page 9: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/9.jpg)
Why did AI choose first-order logic? Provides a declarative substrate
Learn facts, rules from observation and communication Combine and reuse in arbitrary ways
Expressive enough for general-purpose intelligence It provides concise models, essential for learning E.g., rules of chess (32 pieces, 64 squares, ~100 moves)
~100 000 000 000 000 000 000 000 000 000 000 000 000 pages as a state-to-state transition matrix (cf HMMs, automata)
R.B.KB.RPPP..PPP..N..N…..PP….q.pp..Q..n..n..ppp..pppr.b.kb.r ~100 000 pages in propositional logic (cf circuits, graphical models)
WhiteKingOnC4@Move12
9
![Page 10: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/10.jpg)
Why did AI choose first-order logic? Provides a declarative substrate
Learn facts, rules from observation and communication Combine and reuse in arbitrary ways
Expressive enough for general-purpose intelligence It provides concise models, essential for learning E.g., rules of chess (32 pieces, 64 squares, ~100 moves)
~100 000 000 000 000 000 000 000 000 000 000 000 000 pages as a state-to-state transition matrix (cf HMMs, automata)
R.B.KB.RPPP..PPP..N..N…..PP….q.pp..Q..n..n..ppp..pppr.b.kb.r ~100 000 pages in propositional logic (cf circuits, graphical models)
WhiteKingOnC4@Move12 1 page in first-order logic
On(color,piece,x,y,t)
10
![Page 11: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/11.jpg)
11
AI: intelligent systems in the real world
Good Old-Fashioned AI: first-order logic
The world has things
in it!!
![Page 12: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/12.jpg)
12
AI: intelligent systems in the real world
Good Old-Fashioned AI: first-order logic
The world has things
in it!!
The world is
uncertain!!
![Page 13: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/13.jpg)
13
AI: intelligent systems in the real world
Good Old-Fashioned AI: first-order logic
Modern AI: probabilistic graphical models
The world has things
in it!!
The world is
uncertain!!
![Page 14: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/14.jpg)
Bayesian networks
14
Burglary
Alarm
Earthquake
Define distributions on all possible propositional worlds
![Page 15: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/15.jpg)
Bayesian networks
15
Burglary
Alarm
Earthquake
Define distributions on all possible propositional worlds
P(B,E,A) = P(B) P(E) P(A | B, E)
![Page 16: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/16.jpg)
Bayesian networks
16
Burglary
Alarm
Earthquake
Define distributions on all possible propositional worlds
P(B,E,A) = P(B) P(E) P(A | B, E)
0.002
B E P(A|B,E)
F F 0.01
F T 0.4
T F 0.8
T T 0.9
0.003
![Page 17: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/17.jpg)
Bayesian networks
17
Burglary
Alarm
Earthquake
Define distributions on all possible propositional worlds
P(B,E,A) = P(B) P(E) P(A | B, E)
0.0020.003
B E A P(B,E,A)
F F F 0.985
F F T 0.00995
F T F 0.0012
F T T 0.0008
T F F 0.0006
T F T 0.0024
T T F 0.0000006
T T T 0.0000054
B E P(A|B,E)
F F 0.01
F T 0.4
T F 0.8
T T 0.9
![Page 18: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/18.jpg)
Bayesian networks
18
Burglary
Alarm
Earthquake
Define distributions on all possible propositional worlds
P(B,E,A) = P(B) P(E) P(A | B, E)
0.0020.003
B E A P(B,E,A)
F F F 0.985
F F T 0.00995
F T F 0.0012
F T T 0.0008
T F F 0.0006
T F T 0.0024
T T F 0.0000006
T T T 0.0000054
B E P(A|B,E)
F F 0.01
F T 0.4
T F 0.8
T T 0.9
![Page 19: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/19.jpg)
19
AI: intelligent systems in the real world
Good Old-Fashioned AI: first-order logic
Modern AI: probabilistic graphical models
The world has things
in it!!
The world is
uncertain!!
![Page 20: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/20.jpg)
20
AI: intelligent systems in the real world
Good Old-Fashioned AI: first-order logic
Modern AI: probabilistic graphical models
The world has things
in it!!
The world is
uncertain!!
The world is
uncertain!!
![Page 21: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/21.jpg)
21
AI: intelligent systems in the real world
Good Old-Fashioned AI: first-order logic
Modern AI: probabilistic graphical models
The world has things
in it!!
The world has things
in it!!
The world is
uncertain!!
The world is
uncertain!!
![Page 22: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/22.jpg)
22
AI: intelligent systems in the real world
Good Old-Fashioned AI: first-order logic
Modern AI: probabilistic graphical models
A New Dawn for AITM: first-order probabilistic languages
The world has things
in it!!
The world has things
in it!!
The world is
uncertain!!
The world is
uncertain!!
![Page 23: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/23.jpg)
Anil Ananthaswamy, “I, Algorithm: A new dawn for AI,” New Scientist, Jan 29, 2011
![Page 24: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/24.jpg)
“AI is in bloom again … At last, artificial intelligences are thinking along human lines.”
![Page 25: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/25.jpg)
“AI is in bloom again … At last, artificial intelligences are thinking along human lines.”
![Page 26: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/26.jpg)
“AI is in bloom again … At last, artificial intelligences are thinking along human lines.”
“A technique [that] combines the logical underpinnings of the old AI with the power of statistics and probability … is finally starting to disperse the fog of the long AI winter.”
![Page 27: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/27.jpg)
27
First-order probabilistic languages Gaifman [1964]: we can unify logic and probability by
defining distributions over possible worlds that are first-order model structures (objects and relations)
![Page 28: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/28.jpg)
28
First-order probabilistic languages Gaifman [1964]: we can unify logic and probability by
defining distributions over possible worlds that are first-order model structures (objects and relations) Not obvious how to do it – infinitely many parameters??
![Page 29: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/29.jpg)
29
First-order probabilistic languages Gaifman [1964]: we can unify logic and probability by
defining distributions over possible worlds that are first-order model structures (objects and relations) Not obvious how to do it – infinitely many parameters??
Simple idea (1990s): combine logical notation for random variables with Bayes net factorization idea
![Page 30: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/30.jpg)
30
First-order probabilistic languages Gaifman [1964]: we can unify logic and probability by
defining distributions over possible worlds that are first-order model structures (objects and relations) Not obvious how to do it – infinitely many parameters??
Simple idea (1990s): combine logical notation for random variables with Bayes net factorization idea
Burglary(house)
Alarm(house)
Earthquake(Region(house))
![Page 31: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/31.jpg)
Gaifman [1964]: we can unify logic and probability by defining distributions over possible worlds that are first-order model structures (objects and relations) Not obvious how to do it – infinitely many parameters??
Simple idea (1990s): combine logical notation for random variables with Bayes net factorization idea
a
b 31
First-order probabilistic languages
Burglary(house)
Alarm(house)
Earthquake(Region(house))
1 2
3 4 5
![Page 32: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/32.jpg)
32
First-order probabilistic languages Gaifman [1964]: we can unify logic and probability by
defining distributions over possible worlds that are first-order model structures (objects and relations) Not obvious how to do it – infinitely many parameters??
Simple idea (1990s): combine logical notation for random variables with Bayes net factorization idea
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 33: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/33.jpg)
33
First-order probabilistic languages Gaifman [1964]: we can unify logic and probability by
defining distributions over possible worlds that are first-order model structures (objects and relations) Not obvious how to do it – infinitely many parameters??
Simple idea (1990s): combine logical notation for random variables with Bayes net factorization idea
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 34: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/34.jpg)
An important distinction in logic Closed-universe languages assume unique
names and domain closure, i.e., known objects Like Prolog, databases (Herbrand semantics) Poole 93, Sato 97, Koller & Pfeffer 98, De Raedt 00, etc.
Open-universe languages allow uncertainty over the existence and identity of objects
Like full first-order logic BLOG (Milch & Russell 05): declarative OUPM language Probabilistic programming (Pfeffer 03, Goodman et al 08):
distribution on execution traces of stochastic programs
35
![Page 35: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/35.jpg)
36
AI: intelligent systems in the real world
A New Dawn for AITM: first-order probabilistic languages
The world has things in
it and we don’t know what they
are!!
![Page 36: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/36.jpg)
Key idea Given:
An open-universe probability model Evidence from observations
Apply: Bayesian updating Output: beliefs about what objects exist, their
identities, and their interrelations
37
![Page 37: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/37.jpg)
38
A little testGivenBill = Father(William) and Bill = Father(Junior)How many children does Bill have?
![Page 38: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/38.jpg)
39
A little testGivenBill = Father(William) and Bill = Father(Junior)How many children does Bill have?
Closed-universe (Herbrand) semantics:2
![Page 39: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/39.jpg)
40
A little testGivenBill = Father(William) and Bill = Father(Junior)How many children does Bill have?
Closed-universe (Herbrand) semantics:2Open-universe (full first-order) semantics:Between 1 and ∞
![Page 40: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/40.jpg)
Open-universe semanticsPossible worlds for a language with two constant symbols A and B and one relation symbol
41
![Page 41: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/41.jpg)
Open-universe semanticsPossible worlds for a language with two constant symbols A and B and one relation symbol
42but how can we define P on Ω ??
![Page 42: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/42.jpg)
Bayes nets build propositional worlds
43
Burglary
Alarm
Earthquake
![Page 43: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/43.jpg)
Bayes nets build propositional worlds
44
Burglary
Alarm
Earthquake
Burglary
![Page 44: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/44.jpg)
Bayes nets build propositional worlds
45
Burglary
Alarm
Earthquake
Burglarynot Earthquake
![Page 45: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/45.jpg)
Bayes nets build propositional worlds
46
Burglary
Alarm
Earthquake
Burglarynot EarthquakeAlarm
![Page 46: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/46.jpg)
47
Open-universe models in BLOG Construct worlds using two kinds of steps,
proceeding in topological order: Dependency statements: Set the value of a
function or relation on a tuple of (quantified) arguments, conditioned on parent values Alarm(h) ~ CPT[..](Burglary(h), Earthquake(Region(h)))
![Page 47: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/47.jpg)
48
Open-universe models in BLOG Construct worlds using two kinds of steps,
proceeding in topological order: Dependency statements: Set the value of a
function or relation on a tuple of (quantified) arguments, conditioned on parent values Alarm(h) ~ CPT[..](Burglary(h), Earthquake(Region(h)))
Number statements: Add some objects to the world, conditioned on what objects and relations exist so far #GeologicalFaultRegions ~ Uniform{1…10}
![Page 48: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/48.jpg)
Citation information extraction Given: a set of text strings from reference lists:
[Lashkari et al 94] Collaborative Interface Agents, Yezdi Lashkari, Max Metral, and Pattie Maes, Proceedings of the Twelfth National Conference on Articial Intelligence, MIT Press, Cambridge, MA, 1994.
Metral M. Lashkari, Y. and P. Maes. Collaborative interface agents. In Conference of the American Association for Artificial Intelligence, Seattle, WA, August 1994
Decide: What papers and researchers exist For each paper
The real title The real authors The papers it cites
49
![Page 49: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/49.jpg)
50
BLOG model (single-author) #Researcher ~ LogNormal[6.9,2.3]();Name(r) ~ CensusDB_NamePrior();#Paper(Author=r) ~ if Prof(r) then LogNormal[3.5,1.2]() else LogNormal[2.3,1.2]()Title(p) ~ CSPaperDB_TitlePrior();PubCited(c) ~ Uniform({Paper p});Text(c) ~ NoisyCitationGrammar (Name(Author(PubCited(c))), Title(PubCited(c)));
![Page 50: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/50.jpg)
51
BLOG model (single-author) #Researcher ~ LogNormal[6.9,2.3]();Name(r) ~ CensusDB_NamePrior();#Paper(Author=r) ~ if Prof(r) then LogNormal[3.5,1.2]() else LogNormal[2.3,1.2]()Title(p) ~ CSPaperDB_TitlePrior();PubCited(c) ~ Uniform({Paper p});Text(c) ~ NoisyCitationGrammar (Name(Author(PubCited(c))), Title(PubCited(c)));
![Page 51: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/51.jpg)
52
BLOG model (single-author) #Researcher ~ LogNormal[6.9,2.3]();Name(r) ~ CensusDB_NamePrior();#Paper(Author=r) ~ if Prof(r) then LogNormal[3.5,1.2]() else LogNormal[2.3,1.2]()Title(p) ~ CSPaperDB_TitlePrior();PubCited(c) ~ Uniform({Paper p});Text(c) ~ NoisyCitationGrammar (Name(Author(PubCited(c))), Title(PubCited(c)));
![Page 52: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/52.jpg)
53
BLOG model (single-author) #Researcher ~ LogNormal[6.9,2.3]();Name(r) ~ CensusDB_NamePrior();#Paper(Author=r) ~ if Prof(r) then LogNormal[3.5,1.2]() else LogNormal[2.3,1.2]()Title(p) ~ CSPaperDB_TitlePrior();PubCited(c) ~ Uniform({Paper p});Text(c) ~ NoisyCitationGrammar (Name(Author(PubCited(c))), Title(PubCited(c)));
![Page 53: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/53.jpg)
Fraction of citation clusters imperfectly recovered
CiteSeer
ConstraintReasonFaceReinforce0.00
0.05
0.10
0.15
0.20
0.25
Four data sets of ~300-500 citations, referring to ~150-300 papers
![Page 54: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/54.jpg)
Fraction of citation clusters imperfectly recovered
CiteSeerBLOG model
ConstraintReasonFaceReinforce0.00
0.05
0.10
0.15
0.20
0.25
Four data sets of ~300-500 citations, referring to ~150-300 papers
![Page 55: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/55.jpg)
56
BLOG model (single-author) #Researcher ~ LogNormal[6.9,2.3]();Name(r) ~ CensusDB_NamePrior();#Paper(Author=r) ~ if Prof(r) then LogNormal[3.5,1.2]() else LogNormal[2.3,1.2]()Title(p) ~ CSPaperDB_TitlePrior();PubCited(c) ~ Uniform({Paper p});Text(c) ~ NoisyCitationGrammar (Name(Author(PubCited(c))), Title(PubCited(c)));
Origin function
![Page 56: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/56.jpg)
Semantics Objects are defined by type, origin, number:
<Researcher, ,15> <Paper,<Author, <Researcher, ,15>>, 97>
Each basic random variable is a function or predicate symbol indexed by a tuple of objects: Name<<Researcher, ,15>>(ω)
Each possible world ω specifies values for all number variables and basic random variables
Probability of ω is given by the product of conditional probabilities specified in the model
57
![Page 57: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/57.jpg)
58
SemanticsEvery well-formed* BLOG model specifies a
unique proper probability distribution over all possible worlds definable given its vocabulary
* No infinite receding ancestor chains, no conditioned cycles, all expressions finitely evaluable
![Page 58: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/58.jpg)
BLOG Example Library1. PCFG for simple English2. Simplified 3D vision3. Hurricane prediction4. Burglary5. Balls and urns (counting)6. Sybil attack (cybersecurity)7. Students and grades 8. Topic models (LDA)9. Citation information extraction10. Competing workshops11. Galaxy model12. Infinite mixture of Gaussians13. Monopoly (invisible opponent)14. Blackjack
15. Multi-target tracking16. HMM for genetic sequences17. Weather forecasting18. Video background subtraction19. Financial volatility20. Autoregression time series21. Kalman filter22. Infinite-state HMM
59
![Page 59: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/59.jpg)
BLOG Example Library1. PCFG for simple English2. Simplified 3D vision3. Hurricane prediction4. Burglary5. Balls and urns (counting)6. Sybil attack (cybersecurity)7. Students and grades 8. Topic models (LDA)9. Citation information extraction10. Competing workshops11. Galaxy model12. Infinite mixture of Gaussians13. Monopoly (invisible opponent)14. Blackjack
15. Multi-target tracking16. HMM for genetic sequences17. Weather forecasting18. Video background subtraction19. Financial volatility20. Autoregression time series21. Kalman filter22. Infinite-state HMM
60
![Page 60: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/60.jpg)
BLOG Example Library1. PCFG for simple English2. Simplified 3D vision3. Hurricane prediction4. Burglary5. Balls and urns (counting)6. Sybil attack (cybersecurity)7. Students and grades 8. Topic models (LDA)9. Citation information extraction10. Competing workshops11. Galaxy model12. Infinite mixture of Gaussians13. Monopoly (invisible opponent)14. Blackjack
15. Multi-target tracking16. HMM for genetic sequences17. Weather forecasting18. Video background subtraction19. Financial volatility20. Autoregression time series21. Kalman filter22. Infinite-state HMM
61
![Page 61: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/61.jpg)
62
#Aircraft(EntryTime = t) ~ Poisson[λa]();
Exits(a,t) if InFlight(a,t) then ~ Boolean[αe]();
InFlight(a,t)if t < EntryTime(a) then = falseelseif t = EntryTime(a) then = trueelse = (InFlight(a,t-1) & !Exits(a,t-1));
X(a,t)if t = EntryTime(a) then ~ InitState() elseif InFlight(a,t) then ~ Normal[F*X(a,t-1),Σx]();
#Blip(Source=a, Time=t) if InFlight(a,t) then
~ Bernoulli[DetectionProbability(X(a,t))]();
#Blip(Time=t) ~ Poisson[λf]();
Z(b)if Source(b)=null then ~ Uniform[R]()else ~ Normal[H*X(Source(b),Time(b)),Σz]();
Multi-target tracking + data association
![Page 62: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/62.jpg)
BLOG Example Library1. PCFG for simple English2. Simplified 3D vision3. Hurricane prediction4. Burglary5. Balls and urns (counting)6. Sybil attack (cybersecurity)7. Students and grades 8. Topic models (LDA)9. Citation information extraction10. Competing workshops11. Galaxy model12. Infinite mixture of Gaussians13. Monopoly (invisible opponent)14. Blackjack
15. Multi-target tracking16. HMM for genetic sequences17. Weather forecasting18. Video background subtraction19. Financial volatility20. Autoregression time series21. Kalman filter22. Infinite-state HMM
63
![Page 63: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/63.jpg)
BLOG Example Library1. PCFG for simple English2. Simplified 3D vision3. Hurricane prediction4. Burglary5. Balls and urns (counting)6. Sybil attack (cybersecurity)7. Students and grades 8. Topic models (LDA)9. Citation information extraction10. Competing workshops11. Galaxy model12. Infinite mixture of Gaussians13. Monopoly (invisible opponent)14. Blackjack
15. Multi-target tracking16. HMM for genetic sequences17. Weather forecasting18. Video background subtraction19. Financial volatility20. Autoregression time series21. Kalman filter22. Infinite-state HMM
64
![Page 64: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/64.jpg)
65
InferenceTheorem: BLOG inference algorithms (rejection
sampling, importance sampling, MCMC) converge* to correct posteriors for any well-formed model, for any finitely evaluable first-order query
Algorithms dynamically construct finite partial worlds with ground-atom variables directly relevant to query and evidence
![Page 65: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/65.jpg)
MCMC on values
66
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 66: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/66.jpg)
MCMC on values
67
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 67: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/67.jpg)
MCMC on values
68
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 68: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/68.jpg)
MCMC on values
69
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 69: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/69.jpg)
MCMC on values
70
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
Earthquake(Ra)
B(H2)
A(H2)
B(H3)
A(H3)
Earthquake(Rb)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 70: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/70.jpg)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
![Page 71: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/71.jpg)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
![Page 72: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/72.jpg)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
) B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
A(H3
)
Earthquake(R
b)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)B(H1
)
A(H1
)
Earthquake(Ra)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
Earthquake(Rb)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
![Page 73: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/73.jpg)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 74: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/74.jpg)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 75: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/75.jpg)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
![Page 76: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/76.jpg)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H3)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1)
A(H1)
B(H2)
A(H2)
B(H3)
A(H3)
B(H4)
A(H4)
B(H5)
A(H5)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H3
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
B(H1
)
A(H1
)
B(H2
)
A(H2
)
B(H3
)
A(H3
)
B(H4
)
A(H4
)
B(H5
)
A(H5
)
![Page 77: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/77.jpg)
Inference Efficient inference is the bottleneck
Real-world applications use special-purpose inference A substantial engineering problem
Modular design with “plug-in” expert samplers Optimizing compiler technology to reduce overhead Data and process parallelism Incremental query evaluation (cf database systems)
78
![Page 78: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/78.jpg)
79
Application: CTBT
Bans testing of nuclear weapons on earth Allows outside inspection of 1000km2 (18km radius)
183/195 states have signed 159/195 have ratified Need 8 more ratifications including US, China US Senate refused to ratify in 1998
“too hard to monitor”
![Page 79: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/79.jpg)
80
Application: CTBT
Bans testing of nuclear weapons on earth Allows outside inspection of 1000km2 (18km radius)
183/195 states have signed (including Italy) 159/195 have ratified Need 8 more ratifications including US, China US Senate refused to ratify in 1998
“too hard to monitor”
![Page 80: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/80.jpg)
81
Application: CTBT
Bans testing of nuclear weapons on earth Allows outside inspection of 1000km2 (18km radius)
183/195 states have signed (including Italy) 159/195 have ratified (including Italy) Need 8 more ratifications including US, China US Senate refused to ratify in 1998
“too hard to monitor”
![Page 81: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/81.jpg)
82
20534 nuclear explosions, 300K deaths
![Page 82: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/82.jpg)
83
![Page 83: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/83.jpg)
84
287 monitoring stations
![Page 84: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/84.jpg)
85
![Page 85: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/85.jpg)
86
![Page 86: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/86.jpg)
87
![Page 87: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/87.jpg)
Global seismic monitoring Given: continuous waveform measurements
from a global network of seismometer stations Output: a bulletin listing seismic events, with
Time Location (latitude, longitude) Depth Magnitude
88
![Page 88: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/88.jpg)
89
Why is this a hard problem?
~10 000 “detections” per day, 90% false Signals take 15 minutes to several hours to
traverse the earth, so they are all mixed up CTBTO system (GASEL3) developed over 10
years, $100M software plus $1B network Finds 69% of significant events plus about twice as
many spurious (nonexistent) events 16 human analysts correct or discard SEL3 events,
create new events, generate LEB (“ground truth”) Unreliable below magnitude 4 (1kT)
![Page 89: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/89.jpg)
90
![Page 90: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/90.jpg)
91
![Page 91: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/91.jpg)
Very short course in seismology
92
![Page 92: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/92.jpg)
93
![Page 93: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/93.jpg)
94
![Page 94: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/94.jpg)
95
![Page 95: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/95.jpg)
96
![Page 96: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/96.jpg)
97
![Page 97: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/97.jpg)
98
![Page 98: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/98.jpg)
99
![Page 99: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/99.jpg)
100
![Page 100: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/100.jpg)
101
![Page 101: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/101.jpg)
Open-universe model Created a BLOG model describing
Event occurrence Signal propagation Signal detection probability Measurement uncertainty Noise processes producing false detections
Wrote a fast inference algorithm for this model => NET-VISA (network vertically integrated
seismic analysis)
103
![Page 102: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/102.jpg)
104
#SeismicEvents ~ Poisson[T*λe];Time(e) ~ Uniform(0,T)IsEarthQuake(e) ~ Bernoulli(.999);Location(e) ~ if IsEarthQuake(e) then SpatialPrior() else UniformEarthDistribution();Depth(e) ~ if IsEarthQuake(e) then Uniform[0,700] else 0;Magnitude(e) ~ Exponential(log(10));IsDetected(e,p,s) ~ Logistic[weights(s,p)](Magnitude(e), Depth(e), Distance(e,s));#Detections(site = s) ~ Poisson[T*λf(s)];#Detections(event=e, phase=p, station=s) = if IsDetected(e,p,s) then 1 else 0;OnsetTime(a,s) ~ if (event(a) = null) then Uniform[0,T] else Time(event(a)) + GeoTravelTime(Distance(event(a),s),Depth(event(a)),phase(a))
+ Laplace(μt(s), σt(s))
Amplitude(a,s) ~ If (event(a) = null) then NoiseAmplitudeDistribution(s) else AmplitudeModel(Magnitude(event(a)), Distance(event(a),s),Depth(event(a)),phase(a))Azimuth(a,s) ~ If (event(a) = null) then Uniform(0, 360) else GeoAzimuth(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
Slowness(a,s) ~ If (event(a) = null) then Uniform(0,20) else GeoSlowness(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
ObservedPhase(a,s) ~ CategoricalPhaseModel(phase(a))
![Page 103: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/103.jpg)
105
#SeismicEvents ~ Poisson[T*λe];Time(e) ~ Uniform(0,T)IsEarthQuake(e) ~ Bernoulli(.999);Location(e) ~ if IsEarthQuake(e) then SpatialPrior() else UniformEarthDistribution();Depth(e) ~ if IsEarthQuake(e) then Uniform[0,700] else 0;Magnitude(e) ~ Exponential(log(10));IsDetected(e,p,s) ~ Logistic[weights(s,p)](Magnitude(e), Depth(e), Distance(e,s));#Detections(site = s) ~ Poisson[T*λf(s)];#Detections(event=e, phase=p, station=s) = if IsDetected(e,p,s) then 1 else 0;OnsetTime(a,s) ~ if (event(a) = null) then Uniform[0,T] else Time(event(a)) + GeoTravelTime(Distance(event(a),s),Depth(event(a)),phase(a))
+ Laplace(μt(s), σt(s))
Amplitude(a,s) ~ If (event(a) = null) then NoiseAmplitudeDistribution(s) else AmplitudeModel(Magnitude(event(a)), Distance(event(a),s),Depth(event(a)),phase(a))Azimuth(a,s) ~ If (event(a) = null) then Uniform(0, 360) else GeoAzimuth(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
Slowness(a,s) ~ If (event(a) = null) then Uniform(0,20) else GeoSlowness(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
ObservedPhase(a,s) ~ CategoricalPhaseModel(phase(a))
![Page 104: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/104.jpg)
106
#SeismicEvents ~ Poisson[T*λe];Time(e) ~ Uniform(0,T)IsEarthQuake(e) ~ Bernoulli(.999);Location(e) ~ if IsEarthQuake(e) then SpatialPrior() else UniformEarthDistribution();Depth(e) ~ if IsEarthQuake(e) then Uniform[0,700] else 0;Magnitude(e) ~ Exponential(log(10));IsDetected(e,p,s) ~ Logistic[weights(s,p)](Magnitude(e), Depth(e), Distance(e,s));#Detections(site = s) ~ Poisson[T*λf(s)];#Detections(event=e, phase=p, station=s) = if IsDetected(e,p,s) then 1 else 0;OnsetTime(a,s) ~ if (event(a) = null) then Uniform[0,T] else Time(event(a)) + GeoTravelTime(Distance(event(a),s),Depth(event(a)),phase(a))
+ Laplace(μt(s), σt(s))
Amplitude(a,s) ~ If (event(a) = null) then NoiseAmplitudeDistribution(s) else AmplitudeModel(Magnitude(event(a)), Distance(event(a),s),Depth(event(a)),phase(a))Azimuth(a,s) ~ If (event(a) = null) then Uniform(0, 360) else GeoAzimuth(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
Slowness(a,s) ~ If (event(a) = null) then Uniform(0,20) else GeoSlowness(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
ObservedPhase(a,s) ~ CategoricalPhaseModel(phase(a))
![Page 105: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/105.jpg)
Event location prior
109
![Page 106: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/106.jpg)
110
#SeismicEvents ~ Poisson[T*λe];Time(e) ~ Uniform(0,T)IsEarthQuake(e) ~ Bernoulli(.999);Location(e) ~ if IsEarthQuake(e) then SpatialPrior() else UniformEarthDistribution();Depth(e) ~ if IsEarthQuake(e) then Uniform[0,700] else 0;Magnitude(e) ~ Exponential(log(10));IsDetected(e,p,s) ~ Logistic[weights(s,p)](Magnitude(e), Depth(e), Distance(e,s));#Detections(site = s) ~ Poisson[T*λf(s)];#Detections(event=e, phase=p, station=s) = if IsDetected(e,p,s) then 1 else 0;OnsetTime(a,s) ~ if (event(a) = null) then Uniform[0,T] else Time(event(a)) + GeoTravelTime(Distance(event(a),s),Depth(event(a)),phase(a))
+ Laplace(μt(s), σt(s))
Amplitude(a,s) ~ If (event(a) = null) then NoiseAmplitudeDistribution(s) else AmplitudeModel(Magnitude(event(a)), Distance(event(a),s),Depth(event(a)),phase(a))Azimuth(a,s) ~ If (event(a) = null) then Uniform(0, 360) else GeoAzimuth(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
Slowness(a,s) ~ If (event(a) = null) then Uniform(0,20) else GeoSlowness(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
ObservedPhase(a,s) ~ CategoricalPhaseModel(phase(a))
![Page 107: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/107.jpg)
111
Detection probability as a function of distance (station 6, mb 3.5)
P phase S phase
![Page 108: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/108.jpg)
112
#SeismicEvents ~ Poisson[T*λe];Time(e) ~ Uniform(0,T)IsEarthQuake(e) ~ Bernoulli(.999);Location(e) ~ if IsEarthQuake(e) then SpatialPrior() else UniformEarthDistribution();Depth(e) ~ if IsEarthQuake(e) then Uniform[0,700] else 0;Magnitude(e) ~ Exponential(log(10));IsDetected(e,p,s) ~ Logistic[weights(s,p)](Magnitude(e), Depth(e), Distance(e,s));#Detections(site = s) ~ Poisson[T*λf(s)];#Detections(event=e, phase=p, station=s) = if IsDetected(e,p,s) then 1 else 0;OnsetTime(a,s) ~ if (event(a) = null) then Uniform[0,T] else Time(event(a)) + GeoTravelTime(Distance(event(a),s),Depth(event(a)),phase(a))
+ Laplace(μt(s), σt(s))
Amplitude(a,s) ~ If (event(a) = null) then NoiseAmplitudeDistribution(s) else AmplitudeModel(Magnitude(event(a)), Distance(event(a),s),Depth(event(a)),phase(a))Azimuth(a,s) ~ If (event(a) = null) then Uniform(0, 360) else GeoAzimuth(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
Slowness(a,s) ~ If (event(a) = null) then Uniform(0,20) else GeoSlowness(Location(event(a)),Depth(event(a)),phase(a),Site(s)) + Laplace(0,σa(s))
ObservedPhase(a,s) ~ CategoricalPhaseModel(phase(a))
![Page 109: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/109.jpg)
113
Travel-time residual (station 6)
![Page 110: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/110.jpg)
Evaluation 11 weeks of training data, April 6 – June 20, 2009 1 week of validation data, March 22-29, 2009
832 LEB events Evaluated existing UN automated system (GA) and
NET-VISA using LEB as “ground truth”
114
![Page 111: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/111.jpg)
Fraction of events missedExisting UN system
magnitude
![Page 112: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/112.jpg)
116
Fraction of events missed
NET-VISAExisting UN systemNET-VISA
magnitude
![Page 113: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/113.jpg)
Detecting Events Found by Regional Networks
![Page 114: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/114.jpg)
NEIC Event not in LEBLo
g po
ster
ior d
ensi
ty
120
NEIC
NETVISA
![Page 115: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/115.jpg)
North Korea event of 5/25/09
121
Number of associateddetections for event: SEL3 39 NET-VISA 53 LEB 53
50 of 53 detections incommon between LEB and NET-VISA; LEB added 8 by hand
Log
post
erio
r den
sity
NEIC
NETVISA SEL3
LEB
Separate training set 1/4/08-1/4/09Test set 1/5/09-1/26/09
![Page 116: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/116.jpg)
Concluding remarks AI, control theory, operations research, and statistics
tried and failed to find common ground in the 1950s
122
![Page 117: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/117.jpg)
Concluding remarks AI, control theory, operations research, and statistics
tried and failed to find common ground in the 1950s They had no common mathematical formalism!
123
![Page 118: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/118.jpg)
Concluding remarks AI, control theory, operations research, and statistics
tried and failed to find common ground in the 1950s They had no common mathematical formalism!
Robotics and vision separated from AI in the 1970s and 1980s for similar reasons
124
![Page 119: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/119.jpg)
Concluding remarks AI, control theory, operations research, and statistics
tried and failed to find common ground in the 1950s They had no common mathematical formalism!
Robotics and vision separated from AI in the 1970s and 1980s for similar reasons
Bayesian networks and statistical machine learning helped to reintegrate some of these fields in the 1990s
125
![Page 120: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/120.jpg)
Concluding remarks AI, control theory, operations research, and statistics
tried and failed to find common ground in the 1950s They had no common mathematical formalism!
Robotics and vision separated from AI in the 1970s and 1980s for similar reasons
Bayesian networks and statistical machine learning helped to reintegrate some of these fields in the 1990s
Unifying logic and probability may help us to complete the reintegration of reasoning, planning,
perception, and language understanding find the light at the end of the tunnel
126
![Page 121: Unifying logic and probability A new dawn for artificial intelligence?](https://reader034.vdocuments.net/reader034/viewer/2022051700/568161d6550346895dd1dbc8/html5/thumbnails/121.jpg)
Thank you!L'orateur est soutenu par, et cette présentation est donnée sous les auspices de, la Chaire Internationale de Recherche Blaise
Pascal financée par l'État et la Région Île de France, gérée par la Fondation de l'École Normale Supérieure.
Joint work with Brian Milch, David Sontag, Andrey Kolobov, Bhaskara Marthi, Lei Li, Siddharth Srivastava, Nimar Arora, Erik
Sudderth, Paul Kidwell, David Moore, Kevin Mayeda, Steve Myers, Christopher Lin, Tony Dear, Ron Sun, Min Joon Seo
+ help from Ola Dahlman, Ronan LeBras, Lassina Zerbo, Sheila Vaidya, Bob Engdahl, Barbara Romanowicz, Jeff Given, Spilio Spiliopoulos, Elena Tomuta
Research funded by DARPA, CTBTO, and DTRA
Questions?127