Construction of Meaning - DAI

ghostslimAI and Robotics

Feb 23, 2014 (3 years and 5 months ago)

90 views


Martin Tak
ac


Department

of

Computer

Science

University

of

Otago,

New

Zealand

Takáč, M.: Construction of Meanings in
Living and Artificial Agents. Dissertation
thesis, Comenius University, Bratislava,
2007.


Supervisor: Lubica Benuskova

2

Motivation:
W
h
at is it good for
?

Application aspect


Pre
-
defined ontologies are not sufficient in dynamic and open
environments.


It is better to
endow the agents with learning abilities and let them
discover what is relevant and useful for them


=>
developmental approach
to intelligent systems design

3

Motivation:
W
h
at is it good for
?

Philosophy of AI


Can machines understand?


Turing Test

Searl
e’s Chinese Room

Harnad’s Symbol Grounding

Cognitive Science


Better understanding of our own cognition

4

Can machines understand?


Can animals understand?


Can human infants understand?


Depends on the definition of “
understanding
”.


Our approach:
conceive

understanding

in such a way that
the answer is
yes
and look what can we get out of it.

5

Understanding


We say that an agent
understands

its environment,

if it
picks up relevant environmental features and utilizes
them for its goals/survival
.


Situated making of meaning of one’s experience


Semiotics


Umwelt (von Uexkull)


Sign (Peirce)


Understanding is a
gradual phenomenon

in the living
realm ranging from very primitive innate forms to
complex learned human linguistic cognition




Interpretant

(
meaning
)

Obje
c
t

(referent)

Repre
s
entamen

(form)

Sign

6

Key features of meaning


Sensorimotor coupling with the environment


I
ncremental and continuous construction of meaning in
interactions with open and dynamic environment


Collective coordination of individually constructed meanings


[
Takáč, M.: Construction of Meanings in Living and Artificial Agents. In: Trajkovski, G., Collins, S. G. (eds.): Agent
-
Based
Societies: Social and Cultural Interactions, IGI Global, Hershey, PA,
2009
.
]

7

Goal


Propose semantic representation that:


could be i
n
crementally and continuously (re)constructed
from experience/interactions (sensorimotor coupling)


would enable the agent to
understand

its world


causality (prediction of consequences of actions)


planning


inference of intentions/internal states of agents


Do computational implementation and measure the results

8

Roadmap


Semantics of distinguishing criteria


Models of autonomous construction of meanings


By sensorimotor exploration


By social instruction (labelling)


From episodes

9

Roadmap


Semantics of distinguishing criteria


Models of autonomous construction of meanings


By
sensorimotor

exploration


By social instruction (labelling)


From episodes


10


Distinguishing criterion

is

a basic semantic unit
and an
abstrac
t
io
n of the ability to distinguish,
react differentially, understand
(Šefránek, 2002)
.


Semantics of
distinguishing criteria

11


Distinguishing criterion

is

a basic semantic unit
and an
abstrac
t
io
n of the ability to distinguish,
react differentially, understand
(Šefránek, 2002)
.


Neuro
-
b
iologic
al
motiv
ation


Lo
cally tuned detectors

(Balkenius, 1999)


Geometric

representation


Conceptual spaces

(Gärdenfors, 2000)


Semantics of
distinguishing criteria

12

Conceptual spaces



Similarity inversely proportional to distance


Concepts represented by prototypes


learning


a
prototyp
e

computed as centroid of instances


categorization


finding the closest prototype


C
oncept


(
c
onvex)
region

in the space

Metric common for the whole space



獹浭整物捡氠獩浩污物sy

d

13

S
emantics of distinguishing criteria

A distinguishing criterion
r
:


is incrementally constructed from the incoming sequence of
examples of the concept
:

r


{x
1
, …, x
N
}
(learnability)


identifi
es

(
distinguishes
)
instances of the c
oncept:


r
(
x
)


[0,1]
(identification)


auto
-
associatively completes the input
:

r
(
x
)



p
(auto
-
associativity)

14

Distinguishing criteria

Each criterion uses its own metrics with
parameters reflecting statistical properties of
input sample set.

d
2

x

+

)
,
(
2
1
2
)
(
x
p
d
e
x
r



1
-
Σ



All learning starts from scratch, and is
online and incremental!

15

16
/
50

Spe
c
tr
a
l de
c
ompo
sition

of the
c
ovarian
ce

mat
rix







Receptive fields

a
1

a
2

a
1

a
2

a
1

a
2

a
1

a
2

a
1

a
2

.

a
1

a
2

17

Types of distinguishing criteria

“left
_o
f


“big
“,

“blue
“,


tr
iangle


“grew


“house


“a bulldozer pushed the house from the left

“,
“the house fell down


t

t+1

18

Roadmap


Semantics of distinguishing criteria


Models of autonomous construction of meanings


By
sensorimotor

exploration


By social instruction (labelling)


From episodes


19

Roadmap


Semantics of distinguishing criteria


Models of autonomous construction of meanings


By
sensorimotor

exploration


By social instruction (labelling)


From episodes

20


We know how to construct the criterion from its sample set
r


{x
1
, …, x
N
}


Pra
ctical

probl
e
m


to delineate the sample set

(
which criterion
should be fed with the current stimuli?
)



Unsupervised

(
clustering
)


Environmental relevance


By pragmatic feedback


Ecological relevance


By naming (labeling)


Social relevance

Mechanisms of meaning
construction

21


We know how to construct the criterion from its sample set
r


{x
1
, …, x
N
}


Pra
ctical

probl
e
m


to delineate the sample set

(
which criterion
should be fed with the current stimuli?
)



Unsupervised

(
clustering
)


Environmental relevance


By pragmatic feedback


Ecological relevance


By naming (labeling)


Social relevance

Mechanisms of meaning
construction

22

Meaning creation by
sensorimotor

exploration


Environment


Virtual child
, surrounded by objects:
fruits, toys, furniture
.


In discrete time steps, the child performs
random actions
on randomly chosen
objects: trying to
lift

them or
put

them down (with various parameters


force, arm
angle).


Actions performed on objects cause changes of their attribute values. Simple physics
simulated.



Learning


The sensations of the child are in the form of perceptual frames (sets of attribute
-
value pairs) of
objects
,
actions

and
changes

[
x
a
,
x
o
,
x
c
]
.


The child
creates

and
updates

criteria of objects
C
o

, actions
C
a

a
nd

changes

C
c


and
their associations

V


C
a



C
o



C
c


(all sets initially empty).


Objects and actions are grouped to categories by the change. That is,
if an action
leads to the same change on several objects, they will all fall in the same category

and vice versa.


23

Architecture

World

Percep
tion


Agent

Causal module


潢橥瑳
,

a
捴楯cs
,
捯湳c煵敮e敳


Scheduler

Motiva
tion

syst
e
m

needs
,
goals

{ vertices: 3,

posX: 20, posY: 7,

R: 0, G: 0, B: 255 }

Action
repertoire

lift
(


{
force:

10,
angle: 45
} )

24

Meaning creation by
sensorimotor

exploration
-

R
esults


C
au
sal relations


able to predict

consequences of own actions.


Affordances


„Obje
c
t
s too heavy to be lifted
.“


„Obje
c
t
s that cannot be put


down (because they are

already on the ground
).“


Growing sensitivity helpful.

25

Roadmap


Semantics of distinguishing criteria


Models of autonomous construction of meanings


By
sensorimotor

exploration


By social instruction (labelling)


From episodes

26

The agent’s architecture

Pragmatic
s

(ac
tions
,
c
au
s
alit
y
,
goals
, pl
a
n
ning
)

Environment

{ vertices: 3,

posX: 20, posY: 7,

R: 0, G: 0, B: 255 }

Percept
s

Percep
t
i
on


C
oncept
s

Learning

C
ategoriz
ation

big


blue


Language

Child

27

C
ross
-
situational learning


No true homonymy assumption
:


Different words have different senses, even if they share a referent
(in
this case, they denote different
aspects

of the referent).


No true synonymy assumption
:



All referents of a word across multiple situation are considered
instances of the same concept.




周攠浯攠捯nte硴猠潦畳攬瑨攠扥tt敲e捨慮捥c瑨a琠敳獥e瑩t氠
灲潰敲瑩e猠sta礠楮v慲楡n琬睨楬攠畮浰潲慮琠潮敳e睩汬⁶慲y⸠

28

Construction of meaning by
labeling


left_of


„tr
iangle



blue



big
“,


blue
“,


triangle


29

Iter
ated learning


30

Iter
ated learning


31

Iter
ated learning


32

Iter
ated learning


...

33

Construction of meaning by
labeling
-

results


We measured:


similarity of description between teacher and learner


ability to locate the referent(s) of a name


Good meaning similarity between two subsequent generations


Meaning shifts and drift over many generations


Replicator dynamics, more relevant and more general meanings
survive.


Structural meanings more stable.

[ Takáč, M.: Autonomous Construction of Ecologically and Socially Relevant Semantics.
Cognitive Systems Research
9
(4), October 2008, pp. 293
-
311.]

34

Roadmap


Semantics of distinguishing criteria


Models of autonomous construction of meanings


By
sensorimotor

exploration


By social instruction (labelling)


From episodes

35

Roadmap


Semantics of distinguishing criteria


Models of autonomous construction of meanings


By
sensorimotor

exploration


By social instruction (labelling)


From episodes

36

Episodic representation


being learned from
observed/performed actions

Example experiment:


Lattice

5 x 5


4 agent
s

(posX, posY, dir, energy)


10 obje
c
t
s

(posX, posY, nutrition)


Ac
t
i
ons
: move(steps), turn(angle), eat(howMuch)

37

Frame representation of episodes


Rol
e

s
tru
c
t
ure

[
ACT,

SUBJ, OBJ,

SUBJ,

OBJ
]


Example
:

[

ACT
=

{

eat: 1; howMuch:
6

},


SUBJ
=

{

dir:
2
;
@
energy:
10
; posX:
4
; posY:
3

},


OBJ
=

{

nutrition: 129; posX:
3
; posY:
3

},



SUBJ

=

{

dir: 0; @energy:
+
6; posX: 0; posY: 0 }
,




OBJ


=

{

nutrition:
-
6
; posX: 0; posY: 0 }

]


38

Episodic
representation can be
incomplete (partial)


missing roles


missing attributes


because they are internal (private)


due to noise/
stochastic
ity


due to the developmental stage


incompleteness can be used for predictions

39

Re
call from partial episode


[
ACT,

SUBJ
, OBJ,

SUBJ,

OBJ
]


subj
e
c
t
’s abilities

(
what can I do
?)



[
ACT,

SUBJ,
OBJ
,

SUBJ,

OBJ
]


object’s a
f
f
ordanc
es

(
what can be done with it
?)


[
ACT
,

SUBJ, OBJ,

SUBJ,

OBJ
]


verb islands

(
how and upon what to perform the
action
?)


[
ACT,

SUBJ, OBJ,

SUBJ,

OBJ
]


action selection/
pl
a
n
ning
(
how to achieve a desired
change
?
)


40

Requirements


O
pen set of possible attributes


S
tochastic

occurrence of attributes


Learning from observed/performed actions


i
n
c
rement
al


permanent


performance
while learning

&

learning from
performance


Fast learning


reasonable performance after seeing one or
few examples

41

Archite
c
t
ure

Prim
ary layer

Epi
s
odic

layer

[ACT, SUBJ, OBJ,

SUBJ,

OBJ
]

[ACT, SUBJ, OBJ,

SUBJ,

OBJ
]

42

P
rim
ary

layer


t
ransform
s continuous real domain of an attribute to
a vector of

real

[0,1]

activities


covers the real domain with the set of nodes (1
-
dim
detectors), each reacting to a neighborhood of some
real value


neurobiologic
al

motiv
aton

-

prim
a
r
y
sen
s
or
y cort
ic
es

(
localistic

coding
)


qualitatively important landmarks


ap
proxim
ates the distribution of attribute values with
least possible error

43


consist of nodes

{
e
1
,
e
2
, …
e
k
}


epi
s
odic „
memorie
s“


Nodes can be added, refined, merged and forgotten



A node

e
i

:


maintains

N,
A,

i

A:
p
i
,

2
i
, f
i


r
ea
cts to a frame




Episodic layer



N
f
p
x
x
d
i
A
i
i
i
i





2
)
(


1
,
0
},
,...,
{
1


i
n
x
x
x
x

)
(
)
(
x
d
c
e
x
sim





44

Episode
-
based learning
-

Results


Agents able to acquire causal relations (we measured
predictive ability).


Autoassociative recall


potential for simple inferences


Subject’s abilities


Object’s affordances


Prediction


Planning


Inherent
ly

epi
sodic organization of knowledge (implicit
categories of objects, properties, relations and actions)


Prediction of unobservable properties (“empathy” or ToM)

[
Takáč, M., 200
8
.
Developing Episodic Semantics
. In:
Proceedings of AKRR
-
08 (Adaptive Knowledge Representation and
Reasoning)
.
]

45

Mirroring effect
, „empat
hy

,
inference of internal states


A0
sensed (A3


O3):


[
ACT
=

{eat: 1; howMuch: 4; },


SUBJ
=

{dir: 1; posX: 2; posY: 0 },


OBJ
=

{nutrition: 1792; posX: 3; posY: 0 },



SUBJ
=

{dir: 0; posX: 0; posY:
0

},



OBJ
=

{nutrition:
-
4; posX: 0; posY: 0 }

]



A0 r
ecalled
:


[
ACT
=

{eat: 1

(100%);

howMuch:
2

(50%)

}


SUBJ
=

{dir:
0

(50%);

@energy: 40

(46%);

posX:
1

(100%);

posY: 0

(100%)

},


OBJ
=

{nutrition:
1795

(98%);

posX: 3

(100%);

posY: 0

(100%)

},



SUBJ
=

{dir: 0

(100%);

@energy: 2.5

(45%);

posX: 0

(100%);

posY: 0

(100%)

}



OBJ
=

{nutrition:
-
4

(99%);

posX: 0

(100%);

posY: 0

(100%)

}

]


Pragm
. Success
=

0.83

46

Adding communication (future work)


For successful inter
-
agent communication, the meanings should be
mutually coordinated and associated with some signals in a
collectively coherent way.


Speech act as a type of action


Collective dynamics


Pragmatic and contextual language representation


connected to particular states of the speaker (SUBJ) and the hearer
(OBJ), possibly leading to changes of their states (∆SUBJ, ∆OBJ)


prediction/production of different utterances depending on a personal
style and affective state of the speaker, or to infer the internal state of
the speaker from its utterance in some context.


47

Conclusion
-

w
hat we have done


Non
-
anthropocentric
conceptual apparatus

for study of
meanings in different kinds of agents (virtual, embodied,
alive, human...)


Computational representation

of meanings amenable to
autonomous construction


supported by
implemented models
.


Interesting hybrid
computational architecture

that features:


openness in terms of possible attributes and categories or their
gradual change (no catastrophic forgetting)


online learning


from scratch, incremental,
fast
and

permanent


dynamic organization


amenable to analysis of internal structures


48

Conclusion
-

w
hat we have
n’t

done


Cognitive modeling


fit of particular empirical/developmental data


Neuroscience


fit of particular brain structures


Real
-
scale models/applications


complex environments, many agents, noise tolerance


Full
-
blown semantics


abstract meanings, cultural scenarios and many more



… we even haven’t got to language yet…

49

Thank you for your attention!


50