Computational imagination aims to model human imagination by creating artificial agents with intelligence, emotions, and imagination. Imagination is a process of forming semantically linked mental images influenced by perceptions, emotions, context, and prior knowledge. It can be formally represented using visual and linguistic means. Applications could include helping people learn from experience, assisting motor skill learning, aiding older adults' memory, better predicting behavior, and disaster preparation.
2. Motivation
“Imagination is more important than knowledge ..”
Albert Einstein
The power of intelligence cannot be entirely identified with
logical thinking.
“There is no such thing as rational thought ...”
“The solution is to make machines more emotional.”
Marvin Minsky
The Emotion Machine, 2006
3. Understanding Imagination
Imagination is “the faculty or action of forming ideas
or images in the mind.”
Oxford English Dictionary
• Four different conceptions of imagination
as a distinct faculty (Aristotle, Saatre, Kant, Descartes)
as memory or a picture in the mind (Hobbes, Acquinas, Furlong,
Gibson, Hume)
as originality and creativity (Bacon, Kant, Fichte)
imageless imagination (Ryle, White, Wittgenstein)
4. Modelling Imagination
Early computer models agree that imagination uses deep
representations stored in LTM
Description theory – mental images are structured language-based
descriptions
Picture theory – mental images are based on ‘functional pictures’,
and they fade over time
Imagination and emotions (emotional engineering, affective
computing, emotional intelligence)
Imagination in robotics – perceptual activity theory (need to monitor
the environment)
The synthesis of perception and imagination is the core process that
creates new knowledge.
Imagination is essential for innovation and learning.
5. Research Goal
Computational imagination is the science of modelling human
imagination by creating artificial agents with intelligence,
emotions and imagination.
Goal
to study the interplay between cognition, emotion and imagery
to analyse the way perceptions, emotions, prior knowledge and
context influence imagination
to design agents capable of forming concepts and images
6. Research Hypothesis
Based on existing research in connectionism, pictorialism,
descriptionism and perceptual activity theory … and our work on
semantics, ontologies, knowledge structures, context-aware
systems, social agents, computer vision and AI
• Imagination is a process of forming semantically linked mental
images, each representing one or more concepts.
• Imagination can be formally represented using both visual and
linguistic means.
• Imagination is influenced by perceptions, emotions, current context
and prior knowledge.
• The synergy of emotions, imagery and cognition is in the core of
imagination.
7. Application Scenarios
• Imagination gives the ability to look at any situation from a different point of
view, and to mentally explore the past and the future.
• Inventions, buildings, cars, shopping malls, and any kind of business,
started as an idea or concept, and then as a mental image in the mind of
their creators.
• Imagination can be used to unlock creativity,
• Help people/organisations learn from experience by constructing
counterfactual (‘if only’ or ‘what if’) scenarios,
• Learn new motor skills by supplementing motor training with mental
training,
• Improve training by visualising flight training maneuvers,
• Assist older people by, for instance, helping them remember to comply
with medical advice,
• Better predict and understand human behavior in unknown or hostile
environments,
• Prepare for disaster …
8. Challenges
• Descriptive vs. pictorial
• Using both visual and linguistic means
• Using image schemata and conceptual metaphors
• Natural language constructs and ontologies
• Considering prior knowledge (procedural and declarative)
stemming from different sources: innate, interactions with the
environment and culture.
• Considering context: current interactions with the environment,
temporal characteristics and social settings
9. Challenges
• Real-world reasoning
• Highly situated and contextualised
• Extremely unpredictable and entirely personalised
• Alternative ways of reasoning
• Associative reasoning
• Analogical reasoning
• Reasoning under inconsistency (irrational scenarios with contradictions)
• Developing ‘common sense’ in robotics
11. Computational Imagination
• “Imagine”
• A summer holiday (abstract
scenario)
• Your next summer holiday (very
personal and contextualised)
• The first scene is relatively easy
to predict … but the next scenes?
There is NO SCRIPT!
12. Computational Imagination
Imagination is a process of forming semantically linked mental images,
each representing one or more concepts.
Imagination can be formally represented using both visual and
linguistic means.
Imagination is influenced by perceptions, emotions, current context
and prior knowledge.
13. Application Scenarios
To mention just a few …
Help people learn from experience by constructing
counterfactual (‘if only’ or ‘what if’) scenarios,
Learn new motor skills by supplementing motor training with
mental training,
Assist older people by, for instance, helping them
remember to comply with medical advice
Better predict and understand human behavior in unknown
or hostile environments,
Prepare for disaster …