Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs....

13
Computational Creativity and Machine Learning Hannu Toivonen, University of Helsinki, Finland [email protected] 20/11/17 1

Transcript of Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs....

Page 1: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

Computational Creativity and Machine Learning Hannu Toivonen, University of Helsinki, Finland

[email protected]

20/11/17 1

Page 2: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto 20/11/17 2

Page 3: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto 20/11/17 3

Page 4: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto

“Creativity is the ability to come up with ideas or artefacts that are new, surprising, and valuable.”

- Boden 1992

20/11/17 4

Defining creativity

Page 5: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto

Computational creativity is the philosophy, science and engineering of computational systems which, by taking on particular responsibilities, exhibit behaviours that unbiased observers would deem to be creative. - Colton and Wiggins 2012

What is computational creativity?

Page 6: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto

– A purely preprogrammed generative system – only does what it was told to do – has little if any creativity

– Adaptivity or self-determinism –  is necessary to attribute any originality,

responsibility or creative autonomy to a creative system

20/11/17 6

The Opportunity for ML in CC

Page 7: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto

Person, Process, Product, Press Creativity is a complex phenomenon, characterized by many different properties

20/11/17 7

Four Perspectives to Creativity (MacKinnon, 1970; Rhodes, 1961 )

Page 8: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

20/11/17 8

Page 9: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto

– Person perspective: learn skills, develop taste, model emotions or emotional responses, …

– Process perspective: use generative models, solve subtasks related to adaptivity, …

– Product perspective: recognize what is novel, predict the value of artefacts, …

– Press perspective: predict reactions; generate framings

20/11/17 10

Some Possible Uses of Learning in Computational Creativity

Page 10: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto

– Artifact-awareness: what is being created – Generator-awareness: how is it created – Goal-awareness: why is it created –  Interaction-awareness: with whom – Time-awareness: learning and planning – Meta-self-awareness: awareness of being aware Allow reflection and control over the program’s behavior in various respects

20/11/17 11

A Person/Program Perspective: Self-Reflection and Self-Control

Page 11: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto

1.  Concept Extraction: extraction and transformation from an existing but different representation

2.  Concept Induction: learning from examples a)  Concept Learning: supervised, labeled examples b)  Concept Discovery: unsupervised, unlabeled examples

3.  Concept Recycling: creative reuse of existing concepts, e.g.

a)  Concept Mutation: modify one existing concept, e.g., by generalization, specialization, or mutation

b)  Concept Combination: combine many existing concepts 4.  Concept Space Exploration: takes as input a search

space of possible new concepts 20/11/17 12

A Typology of Concept Creation

Page 12: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

www.helsinki.fi/yliopisto 20/11/17 13

Machine Learning Problems vs. Computational Creativity

Machine learning problems Computational Creativity problems

Well - specified ( e.g ., ” induce a classifier ”)

Ill - defined , open - ended ( e.g . ” write a poem ”)

Have obvious and objective success criteria (e.g. classification accuracy)

Have subjective and non - explicit criteria ( e.g . when is a poem good ?)

Success can be measured with relative ease ( e.g . evaluate on test set)

Evaluation cannot be computed easily ( e.g . ask subjects to evaluate )

Page 13: Computational Creativity and Machine Learning · 2018-04-16 · Machine Learning Problems vs. Computational Creativity Machine learning problems Computational Creativity problems

Thank you!

20/11/17 14