Timo Honkela: An Introduction to Artificial Intelligence
1. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
Timo Honkela
Modeling Meaning and Knowledge
1 Feb 2016
timo.honkela@helsinki.fi
An Introduction to
Artificial Intelligence
2. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
AI and humanities?
● What the relationship between humanities and
artificial intelligence?
● Many areas of AI are directly related to areas
in humanities
– Linguistics and natural language processing
– Epistemology and knowledge representation
● AI methods can be useful as analytical tools in
digital humanities
● The use of AI raises many ethical questions
3. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
Artificial Intelligence:
Some early modern history
● Alan Turing (1912 – 1954)
– Turing machine: theoretical framework for
symbolic computation
– Turing test:
● Dartmouth Conference in 1956
– e.g. Marvin Minsky, Allen Newell, Arthur Samuel,
and Herbert Simon participated
4. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
https://en.wikipedia.org/wiki/Alan_Turing
Turing proposed an
experiment in which
a human interrogator
would have a written
conversation with two
participants, one
human and another
computer-based. If the
interrogator could not
tell who is human and
which machine,
artificial intelligence
is reached.
Turing test
5. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
https://en.wikipedia.org/wiki/Alan_Turing
Turing proposed an
experiment in which
a human interrogator
would have a written
conversation with two
participants, one
human and another
computer-based. If the
interrogator could not
tell who is human and
which machine,
artificial intelligence
is reached.
Turing test
Intelligence is defined here
to be strongly associated
with linguistics abilities.
6. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
https://en.wikipedia.org/w/index.php?title=Artificial_intelligence
http://www.newschool.edu/nssr/het/profiles/simon.htm
https://en.wikipedia.org/wiki/Marvin_Minsky
Herbert Simon predicted that
"machines will be capable,
within twenty years, of doing
any work a man can do".
Marvin Minsky agreed: "within
a generation ... the problem of
creating 'artificial intelligence'
will substantially be solved".
7. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
SPEECH RECOGNITION
SPEECH SYNTHESIS
NATURAL LANGUAGE PROCESSING PATTERN RECOGNITIO
REASONING
KNOWLEDGE REPRESENTATION
PROBLEM SOLVING
MACHINE TRANSLATION
SOCIAL AI
MODELING EMOTIONS
MACHINE VISION
PLANNING
GAMES
ARTIFICIAL LIFE
Some central areas of artificial intelligence research
8. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
https://books.google.com/ngrams/
Some AI related terms in Google Books
Neural networks
Artificial intelligence
Semantic web
Machine learning
Knowledge representation
Natural language processing
Machine translation
9. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
https://books.google.com/ngrams/
“Machine Translation” in Google Books
10. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
World Knowledge
● For a long time it was a considerable problem
for AI systems to “possess” enough knowledge
to be able to dead with real world problems
and tasks – cf. e.g. Winograd's SHRDLU
● This problem has been alleviated by progress
in two fronts
– Large Semantic Web systems
– Applying machine learning on big data collections
https://en.wikipedia.org/wiki/SHRDLU
11. Timo Honkela, Modeling Meaning and Knowledge, 1.2.2016
Machine Learning
● Theoretically founded approaches
– Theory-driven logic-based approaches: inductive inference
– Data-driven approaches: Based on a) probability theory and
statistics, b) Based on information theory
● Biologically inspired approachs
– Artificial neural networks
– Evolutionary computing
● Machine learning as a means (data science,
engineering) or as an end (biological, cognitive,
social etc. adaptive systems)