403 Presentation

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 20

Modeling Consciousness:

A Review
Travis Beam, Alayna Daniels, Jason Lim,
Charles Parsons, Haneul Ryou
● Introduction
○ Overview of models
● Similarities
● Differences
● Competing Expectations
Outline ● Discussion
○ Implications
○ Future Areas
○ Missing Topics
● Works Cited
Main Fields: Major Computational Challenges:
Correlates:
● Global workspaces ● Architectural specialization ● Understanding the basis of
consciousness
● Information integration ● Properties of psychologically
representations
● Internal self-models
● Properties of specific
● Higher-level processes
representations

● Attention

Introduction to Artificial
Consciousness
The Global Workspace

● The underlying phenomenon ● Represents a temporary ● Creates a global/exclusive


of a wide array of metastable state of global availability for specific stimuli
computational correlate activity to be experienced as
candidates consciousness
COALIA Brain-Scale ● Simulates brain activity
associated with
Electrophysiological consciousness
activity Model
● Uses local neuron
population to capture
dynamics of larger
assemblies and neural mass
modeling

● Focuses on simulating both


cortico-cortical and
cortico-thalamic activity
● Applies to a vast array of
ACT-R Cognitive behaviors and cast theories
on the components of
Architecture Model cognition

● Models Cognitive processes


through modules which map
onto regions of the brain that
act in parallel but are
processed serially

● Use “chunks,” and buffers


to create a bottle next for
information transfer
The Forward Model
● Develops a model of the system to be
controlled and learns to control the system

● Consist of two interconnected networks


forming the basis of meta-representations

● Entirely dependent on action and model


through integrated functional account
Similarities
Self Sustaining

● Critical aspects of consciousness include...


○ Self awareness
○ Predict future outcomes (Ascoli, 2005)
● Models are able to carry forward self-produced actions
○ Otherwise distinguish the ‘self’ from ‘others’
○ Need initial training

*limitations to modeling consciousness because no models are


perfect without intervention
All-Encompassing Model

● Incorporate all parts of the brain to represent a single, global network of


neurons
○ Global Workspace
○ COALIA → neural mass modeling
○ ACT-R → ‘chunks’ in parallel, processed serially
○ Forward-Facing → two processes work closely together

*no single portion or process can model consciousness


Differences
COALIA vs Global Workspace

● Differences in a physiological based approach

● COALIA: Focus is on microcircuitry


○ Based on neurobiological studies
○ Dynamics of neuronal assemblies

● Global Workspace: Focus is on whole brain activation


○ Establish hierarchical signaling pathways
■ Bottom-up and top-down
ACT-R

● Not as focused on physiologically based modeling


○ Computational model
○ Focuses on the brain regions

● Model brain as a whole


○ Modules for brain regions
○ Perception, speech, memory, etc.

● Metabolic activity
○ Estimated blood supply
○ Brain activity related to behavior
Competing
Expectations
1. Behavioral
2. Physiological
3. Computational
Competing Expectations: Behavioral

● Expectation 1: Central function is to


predict and have adaptable control to
behavior (Cleereman, p.4)

● Expectation 2: separate out cognitive capacities


○ Good behavioral predictors for manual, retrieval, imaginal, procedural
tasks. Bad for visual.
○ “In fact, this approach might become a trend that, so we and others think, has the potential to ultimately revolutionize the field
—once entry-level barriers to complex architectural modeling tools such as ACT-R break away” (Dimov et al., p. 874)
Competing Expectations: Physiological

● Expectation 3: There is a potential to create an objectively


physiological model of consciousness
○ Neurological model neurons, model ‘layers’ as cortical ‘layers’.
(Dehaene et al., p.8521)

● Expectation 4: Computational
models of consciousness provide
pathological implications (Bensaid et al.)
Competing Expectations: Computational

● Expectation 5: Modeling consciousness is


worth doing through machine and artificial
consciousness.
○ “It will remain very difficult to create
artifacts that truly model or support
analogous artificial conscious states...Once
an improved understanding occurs, advances
will probably be very rapid [possibly
inevitable].” (Reggia, p. 129)
Implication: Missing Part
● Limitation

● Assumptions & simplifications


○ Definition
○ Reducing/simplifying physiology
○ Excluding human subjectivity
○ Reliant on model components

● Consciousness as a whole?
Implication: Future Direction & Conclusion

Incorporation of basic human components i.e.


physiological grounding

Reducing simplification and assumptions and improving


model architecture

Integrating different models

Transcend
Works Cited
Bensaid, S., Modolo, J., Merlet, I., Wendling, F., & Benquet, P. (2019). COALIA: A computational model of HUMAN EEG for consciousness

research. Frontiers in Systems Neuroscience, 13, 1-18. doi:10.3389/fnsys.2019.00059

Cleeremans, A. (2005). Computational Correlates of Consciousness. Progress in Brain Research, 150, 81-98.

doi:10.1016/s0079-6123(05)50007-4

Dehaene, S., Sergent, C., & Changeux, J. (2003). A neuronal network model linking subjective reports and objective physiological data

during conscious perception. Proceedings of the National Academy of Sciences, 100(14), 8520-8525. doi:10.1073/pnas.1332574100

Dimov, C., Khader, P. H., Marewski, J. N., & Pachur, T. (2019). How to model the Neurocognitive dynamics of decision making: A

Methodological primer with ACT-R. Behavior Research Methods, 52(2), 857-880. doi:10.3758/s13428-019-01286-2

Reggia, J. A. (2013). The rise of machine consciousness: Studying consciousness with computational models. Neural Networks, 44, 112-131.

doi:10.1016/j.neunet.2013.03.011

You might also like