8.7 Artificial Intelligence

Transcription

8.7 Artificial Intelligence
Hva er “Kunstig intelligens” – 1
Kunstig intelligens
- en kort introduksjon
“AI = Things that make you go WOW!”
eller…??
Agnar Aamodt
Institutt for datateknikk og informasjonsvitenskap
Gruppe for Intelligente Systemer
NTNU
A. Aamodt, NTNU-IDI
Well, somewhat more sober although more dull:
•  AI enables systems to perform tasks in ways that would be called intelligent
if done by humans.
•  AI enables complex problem solving and interaction beyond what other
(non-AI) methods do.
A. Aamodt, NTNU-IDI
8.7 Artificial Intelligence
Hva er “Kunstig intelligens” – 2
INFORMATIKK
STUDIET AV
RELATERT
INTELLIGENTE
REALISERING AV DATASYSTEMER SOM KAN
SYSTEMER
TIL KOMPUTASJONELLE
SIES Å OPPVISE INTELLIGENT ADFERD
- DVS . ' SMARTERE ' SYSTEMER
er delfelt av
PROSESSER
er koblet via empirisk vitenskapelig metode
har
vitenskapelig
perspektiv
har
teknologisk
perspektiv
KUNSTIG INTELLIGENS (AI)
bygger bl.a. på
har metoder
har metoder
MATEMATIKK
SYMBOLORIENTERTE METODER
(KUNNSKAPSBASERTE METODER)
FILOSOFI
KOGNITIV
PSYKOLOGI
BIOLOGI
A. Aamodt, NTNU-IDI
SUBSYMBOLSKE METODER
(BIO-INSPIRERTE METODER)
!  AI is a group of related technologies used to develop
software and machines that emulate human qualities such
as learning, reasoning, communicating, seeing, and hearing
!  Areas include:
!  Expert systems
!  Natural language processing
!  Intelligent agents
!  Pattern recognition
!  Fuzzy logic
!  Virtual reality and simulation devices
!  Robotics
© 2011 The McGraw-Hill Companies,
Inc. All rights reserved.
A. Aamodt, NTNU-IDI
KUNNSKAPSBASERTE METODER
- UTVIKLINGSTRENDER
Core idea
“AI = Representation + Search”
Kontroll-kunnskap
•  The concept of search plays an important role in science and
engineering
Heuristiske
regler
•  Any problem whatsoever can be seen as a search for
“the right answer” – or at least “a good answer”
Dyp kunnskap
•  This search can be viewed to take place in a “problem space”,
which constrains the search through its representation
A. Aamodt, NTNU-IDI
Spesifikke
case
Fra generell kunnskap til situasjons-spesifikke case
(f.eks. CYRUS, PROTOS)
- case-basert resonnering
A. Aamodt, NTNU-IDI
Hypen rundt “A.I.”
KUNNSKAPSBASERTE METODER
“… in three to eight years we will have a machine
with the general intelligence of an average
human being. I mean a machine that will be
able to read Shakespeare, grease a car, play
office politics, tell a joke, and have a fight. At
that point, the machine will start to educate
itself with fantastic speed. In a few months it
will be at genius level and a few months after
that its powers will be incalculable.”
Marvin Minsky — i 1970 !
- er karakterisert ved at
•
kunnskap representeres eksplisitt - modelleres - i
form av navngitte begreper og relasjoner mellom
disse
•
begreps-termene kalles symboler, da de
symboliserer - står for - begreper i den virkelige
verden (objekter, hendelser, prosesser, ...)
•
symbol-strukturene som derved bygges opp,
prosesseres av tolknings- og slutningsmetoder
(inferensmetoder) slik at ny informasjon kan
utledes, konklusjoner kan trekkes, og ny kunnskap
kan læres
•
representasjonen av kunnskapen er adskilt fra
metodene som prosesser den
- og er kjernemetodene innen kunnskapsteknologi
for å utvikle kunnskapsbaserte systemer
.
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
KUNNSKAPSBASERTE METODER
Representasjon av det som for oss
mennesker fortoner seg som kunnskap, i
form av symbolstrukturer og inferensmetoder i et datasystem, og på en måte
som gjør systemet istand til å tolke
symbolene og utifra dette trekke meningsfylte slutninger, gjør at symbolstrukturene
kan betraktes som kunnskap for systemet.
Dette danner basis for det 'kunnskapsbaserte paradigmet'
innen AI, og for rekken av kunnskapsbaserte systemer som er
utviklet og i daglig bruk.
Refereres ofte til som ”GOFAI” – Good Old-Fashioned AI,for å skille denne typen
metoder fra ikke-symbolske, eller sub-symbolske tilnærminger til AI.
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Fuzzy Logic Reasoning
A computer chess success
“Everything is matter of degree”
Critical Serious
Minor
Uninjured
1.0
Kunnskap beskrives som regler som
det gjøres ”Fuzzy reasoning” på:
http://www.youtube.com/watch?v=NJarxpYyoFI
Near
Death
0.0
Health
If (range is Melee AND health is Uninjured)
AND rating is NOT Hard THEN action is Attack !
If ( range is NOT Melee AND health is
Uninjured) ! THEN action is Donothing!
If (range is Melee AND health is NOT
Uninjured) AND rating is NOT Wimp THEN !
!
!
!
!
!
!
!action is Flee!
Easy
1.0
A. Aamodt, NTNU-IDI
Hard
Wimp
0.0
“qualitative description used in everyday language”
Moderate
Rating
A. Aamodt, NTNU-IDI
En kritikk av kunnskapsbaserte metoder
A Jeopardy success: Watson
•
Mennesker representerer sin kunnskap i maskinen,
den kan defor ikke bli maskinens .
Nyttig kritikk, men satt på spissen.
Berører kunnskapsrepresentasjons-problemet generelt.
-
Konsekvenser for interaktive rådgivingssystemer :
Utviklingsmetoder der mindre vekt legges på datamaskinen
som problemløser, mer på samspillet mellom maskin og bruker.
En utfordring er at systemet må ha eller bygge opp kunnskap
om interaksjons-omgivelsen.
-
Konsekvenser for maskinlæring :
Større grad av erfaringsbasert læring, der systemene lærer
gradvis ved å delta i løsningen av reelle problemer i det miljø
de virker.
For (morgen)dagens ‘ekspertsystemer’:
Fra 'tradisjonelle' problemløsende systemer
til aktive, interaktive, og adaptive beslutningsstøtte-systemer
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Nevrale Nett!
Biologisk bakgrunn: Et nettverk av nerveceller!
SUBSYMBOLSKE METODER (eks.)
•
Cell !
body!
NEVRALE NETT
Distribuert representasjon i et nettverk av noder koblet
sammen via lenker (konneksjonisme).
Ingen eksplisitt begreper, direkte representasjon av
sammenhenger.
•
GENETISKE ALGORITMER
Representasjon i form av bit-strenger som maniplueres vha.
genetiske operatorer.
Bit-strengene kan representere begreper, men ikke
nødvendigvis.
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Bio-inspired methods
Swarm Intelligence
•  The signal feature of life is not the carbon-based substrate...(but)...that the
local dynamics of a set of interacting entities (e.g. molecules, cells, etc.)
supports an emergent set of global dynamical structures which stabilize
themselves by setting the boundary conditions within which the local
dynamics operates (Charles Taylor, biologist, UCLA)
A. Aamodt, NTNU-IDI
•  Follow Trail
•  Find Food
• Make Trail
A. Aamodt, NTNU-IDI
Darwinian Evolution
Evolutionary Algorithms
Physiological, Behavioral
Phenotypes
Natural Selection
Parameters,
Code,
Neural Nets,
Rules
Ptypes
Semantic
Performance Test
P,C,N,R
Reproduction
Sex
Morphogenesis
R &M
Translate
Generate
Genotypes
Recombination
& Mutation
Gtypes
Bit Strings
Genetic
A. Aamodt, NTNU-IDI
Syntactic
A. Aamodt, NTNU-IDI
Recombination
& Mutation
Bits
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
8.7 Artificial Intelligence
Turing-testen"
!  Turing Test
!  In 1950 Allen Turing predicted computers would
eventually be able to mimic human thinking
!  Turing test determines whether the computer is human
A. Aamodt, NTNU-IDI
!  Judge is in another location and doesn’t see the computer
!  Judge converses via a computer terminal with two entities:
one a person and one a computer
!  Judge must determine who is the person and who the
computer
!  If the computer can fool the judge, it is said to be intelligent
!  No computer system has yet passed the Turing test
© 2011 The McGraw-Hill Companies,
Inc. All rights reserved.
A. Aamodt, NTNU-IDI
8.7 Artificial Intelligence
!  Weak vs. Strong AI
!  Weak AI
8.7 Artificial Intelligence
!  Strong AI
!  Cyc approach to strong AI
!  Computers can be programmed to simulate human cognition
!  A database in Austin, TX that holds about 1.4 million basic
truths
!  Plan is that Cyc will automatically make human-like
assumptions
!  Hope is that Cyc will learn on its own
!  Strong AI
!  Computers can think on a level that is equal to or better than
humans and can also achieve consciousness
!  Cog approach to strong AI
!  MIT project that is a humanoid robot with sensory systems
!  Tries to identify and search for patterns instead of following
rules and facts
© 2011 The McGraw-Hill Companies,
Inc. All rights reserved.
A. Aamodt, NTNU-IDI
© 2011 The McGraw-Hill Companies,
Inc. All rights reserved.
A. Aamodt, NTNU-IDI
8.7 Artificial Intelligence
!  Smarter-Than-Human Computers
!  “The Singularity”
!  A moment when humans would have created self-aware,
smarter-than-human machines capable of designing
computers and robots that are better than humans can design
today
!  Also may involve transferring the contents of human brains
and thought processes into a computing environment
© 2011 The McGraw-Hill Companies,
Inc. All rights reserved.
A. Aamodt, NTNU-IDI
8.7 Artificial Intelligence
!  Ethics in A.I.
!  Computer software is subtly shaped by the ethical
judgments and assumptions of its creators.
!  Will humans lose control of computer systems?
!  There is no such thing as completely value-free
technology.
© 2011 The McGraw-Hill Companies,
Inc. All rights reserved.
A. Aamodt, NTNU-IDI
NTNU
Eksempler på aktiviteter
NTNU
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Eksempler på master-oppgaver
Eksempler på master-oppgaver
Bilde- og/eller Video-analyse
(Her: Segmentere bilder av fisk i Mauritius)
Improved game AI through case-based and statistical reasoning
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Eksempler på master-oppgaver
Eksempler på master-oppgaver
Intelligent Hardware
Robots (pictured) that
interact with either a real
or simulated other robot.
Within our PUCKER
system, researchers and
students can easily test
their AI control strategies
on this type of robot (epucks).
Today’s hardware technologies, especially Field programmable
Gate Arrays (FPGAs), provide many possibilities for the
creation of intelligent Hardware - that is AI techniques
embedded in hardware.
Such embedding may be for the purpose of speed-up of a given AI
technique for perhaps real-time application requirements or for
the purpose of creating hardware circuits, applying bioinspired techniques as the design technique.
The latter is known as the field of Evolvable Hardware and
includes applications in today’s technology and approaches to
achieve computation in tomorrow’s technology. Application
areas range from Vision, art to electronic circuits.
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Eksempler på master-oppgaver
Eksempler på master-oppgaver
Språkteknologi - maskinoversetting
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Eksempler på master-oppgaver
Eksempler på master-oppgaver
Data mining and Decision support in Fish Farming
Textual CBR.
Discovery of causal relations in incident reports
•  An incident report (i.e., a 'textual case') describes how a problem unfolds. That is, the story
starts with less important 'symptoms'/evidence which, in turn, triggers/causes more serious
ones, and this chain of evidence ends up with an undesired, anomalous event. It is
important to identify the events when they are small, and discover the causal mechanisms
underlying the chain of events.
•  Use of eye-tracking in the selection of important features in a text and determining how
important they are - the latter is called 'weighting’. This in cooperation with people at
Dragvoll.
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Eksempler på master-oppgaver
Reducing unwanted downtime in oil drilling
•  One day of unwanted downtime on this rig
means increased cost of 1,6 MNOK for the
ongoing drilling operation.
•  Providing the relevant experience and
getting the right information precisely
when needed will reduce unwanted
operational downtime.
•  The result is a more reliable drilling
process, reduced drilling costs, and
increased productivity.
A. Aamodt, NTNU-IDI
IDIs AI-gruppe har deltatt i etablering av tre
spin-off selskaper:
-  LingIT AS
- naturlig språk tolkning og dialogsystemer
-  Trollhetta AS
-  bildeanalyse og beslutningsstøtte
-  Verdande Technology AS
- erfarings-lagring og aktiv gjenbruk,
primært innen oljeboring
A. Aamodt, NTNU-IDI
Eksempler på master-oppgaver
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI
Artificial Intelligence (AI)
- represents a set of computer science methods
aimed at intelligent behaviour of computer systems
- covers a lot of methods and application areas
- is interesting, useful, and fun
A useful link to all of AI: http://www.aaai.org/aitopics
A. Aamodt, NTNU-IDI
A. Aamodt, NTNU-IDI