#LyX 2.3 created this file. For more info see http://www.lyx.org/
\lyxformat 544
\begin_document
\begin_header
\save_transient_properties true
\origin unavailable
\textclass llncs
\begin_preamble
\usepackage{url}
\usepackage{slashed}
\end_preamble
\use_default_options true
\maintain_unincluded_children false
\language english
\language_package default
\inputencoding utf8
\fontencoding global
\font_roman "times" "default"
\font_sans "helvet" "default"
\font_typewriter "cmtt" "default"
\font_math "auto" "auto"
\font_default_family default
\use_non_tex_fonts false
\font_sc false
\font_osf false
\font_sf_scale 100 100
\font_tt_scale 100 100
\use_microtype false
\use_dash_ligatures false
\graphics default
\default_output_format default
\output_sync 0
\bibtex_command default
\index_command default
\paperfontsize default
\spacing single
\use_hyperref true
\pdf_bookmarks true
\pdf_bookmarksnumbered false
\pdf_bookmarksopen false
\pdf_bookmarksopenlevel 1
\pdf_breaklinks true
\pdf_pdfborder true
\pdf_colorlinks true
\pdf_backref false
\pdf_pdfusetitle true
\papersize default
\use_geometry false
\use_package amsmath 2
\use_package amssymb 2
\use_package cancel 1
\use_package esint 0
\use_package mathdots 1
\use_package mathtools 1
\use_package mhchem 0
\use_package stackrel 1
\use_package stmaryrd 1
\use_package undertilde 1
\cite_engine basic
\cite_engine_type default
\biblio_style splncs04
\use_bibtopic false
\use_indices false
\paperorientation portrait
\suppress_date false
\justification true
\use_refstyle 0
\use_minted 0
\branch long-version
\selected 0
\filename_suffix 0
\color #faf0e6
\end_branch
\index Index
\shortcut idx
\color #008000
\end_index
\leftmargin 11page%
\topmargin 8pheight%
\rightmargin 11page%
\bottommargin 10pheight%
\secnumdepth 3
\tocdepth 3
\paragraph_separation indent
\paragraph_indentation default
\is_math_indent 0
\math_numbering_side default
\quotes_style english
\dynamic_quotes 0
\papercolumns 1
\papersides 1
\paperpagestyle default
\listings_params "basicstyle={\ttfamily},basewidth={0.45em}"
\tracking_changes false
\output_changes false
\html_math_output 0
\html_css_as_file 0
\html_be_strict false
\end_header
\begin_body
\begin_layout Title
Agency and Sensorimotor Systems
\end_layout
\begin_layout Subtitle
Response to a Call for Research Proposals, California Institute for Machine
Consciousness (CIMC)
\end_layout
\begin_layout Author
Linas Vepštas
\begin_inset ERT
status open
\begin_layout Plain Layout
\backslash
orcidID{ORCID
\backslash
hspace{0.75em} 0000-0002-2557-740X}
\end_layout
\end_inset
\end_layout
\begin_layout Institute
BrainyBlaze Dynamics, OpenCog Foundation
\begin_inset ERT
status open
\begin_layout Plain Layout
\backslash
email{<linasvepstas@gmail.com>}
\end_layout
\end_inset
\end_layout
\begin_layout Abstract
The core idea of consciousness as a form of embodied cognition is popular
and widespread.
Yet an abstract mathematical description of how this might work, anchored
to first principles in physics, is absent.
A sensory system
\begin_inset Quotes eld
\end_inset
perceives
\begin_inset Quotes erd
\end_inset
the world: but what is it that is perceived? It can't just be some
\begin_inset Quotes eld
\end_inset
data
\begin_inset Quotes erd
\end_inset
or
\begin_inset Quotes eld
\end_inset
message
\begin_inset Quotes erd
\end_inset
, as that seems to require a data–definition format and some homunculus
to define it.
Perhaps sensation is movement of structural relationships across a barrier
or boundary between the agent's finite and limited self/inside and the
unbounded outside of the external world.
This requires the barrier only to limit what gets across.
What is the mathematical description of such a barrier? Perhaps the
\begin_inset Quotes eld
\end_inset
self
\begin_inset Quotes erd
\end_inset
is an accretion of structure inside an agent.
How can this accretion be formally described? Jellyfish seem to be aware
of
\begin_inset Quotes eld
\end_inset
self
\begin_inset Quotes erd
\end_inset
and
\begin_inset Quotes eld
\end_inset
other
\begin_inset Quotes erd
\end_inset
; discussions of consciousness often presume such awareness.
Is there an algebraic formulation for such (self-)awareness? Suppose such
a formulation can be created; is there anything within it that has the
form of
\begin_inset Quotes eld
\end_inset
qualia
\begin_inset Quotes erd
\end_inset
, or is it mechanical and dead inside?
\end_layout
\begin_layout Section*
Research Objectives
\end_layout
\begin_layout Standard
Computation requires the articulation of a system using algorithmic language.
Algorithms require precise structural definitions.
Structural definitions are generically algebraic.
Foundations for algebra include set theory, category theory and topoi.
Thus, a
\begin_inset Quotes eld
\end_inset
computational perspective on consciousness
\begin_inset Quotes erd
\end_inset
fundamentally requires (seems to require) a reductionist description of
the structure of the exterior world, the barrier that separates it from
the interior world, and how that interior interacts and moves through the
external world.
The interior itself may have hierarchical structure: the inside of a biological
creature is not a uniform puree of pancake batter, but also contains substructu
res.
It seems this recursive structuring continues
\begin_inset Quotes eld
\end_inset
all the way down
\begin_inset Quotes erd
\end_inset
, at least as far as biochemistry,
\begin_inset Foot
status collapsed
\begin_layout Plain Layout
See the review article:
\begin_inset Quotes eld
\end_inset
All intelligence is collective intelligence
\begin_inset Quotes erd
\end_inset
, Falandays
\emph on
etal
\emph default
.
Journal of Multiscale Neuroscience Vol 2 no 1 169-191 (2023) https://researchpo
rtal.helsinki.fi/en/publications/all-intelligence-is-collective-intelligence
\end_layout
\end_inset
and perhaps much farther.
\begin_inset Foot
status collapsed
\begin_layout Plain Layout
See
\begin_inset Quotes eld
\end_inset
Ingressing Minds: Causal Patterns Beyond Genetics and Environment in Natural,
Synthetic, and Hybrid Embodiments
\begin_inset Quotes erd
\end_inset
, Michael Levin https://osf.io/preprints/psyarxiv/5g2xj_v1
\end_layout
\end_inset
\end_layout
\begin_layout Standard
The goal of the research is to articulate what it means for there to be
an inside, an outside, and a permeable boundary between the two.
Does this boundary resemble a domain wall in a ferromagnetic system (
\emph on
a la
\emph default
Ising model)? Just as petroleum does not flow through rock, until there
are enough holes/fractures in the rock allowing it to flow, exhibiting
a phase transition at a critical fracture–ratio, then perhaps intelligence
is similar: absent, until there is a critical level of organization? Many
systems in nature exhibit self–organized criticality, and recent neuroscience
results provide some indication that human intelligence operates at the
point of self–organized criticality, with long neurons modulating the avalanchi
ng behavior that is characteristic of self–organized systems.
\begin_inset Foot
status collapsed
\begin_layout Plain Layout
See
\begin_inset Quotes eld
\end_inset
Complex harmonics reveal low–dimensional manifolds of critical brain dynamics.
\begin_inset Quotes erd
\end_inset
Deco,
\emph on
etal.
\emph default
https://journals.aps.org/pre/abstract/10.1103/PhysRevE.111.014410
\end_layout
\end_inset
\end_layout
\begin_layout Standard
One can imagine that it is easy to write a mathematical description of inside,
outside and a barrier between them.
It is hardly clear how to take such a system, and have it drive itself
into a self–organized critical point, exhibiting any of the basic phenomena
associated with intelligence.
\end_layout
\begin_layout Section*
Methodology
\end_layout
\begin_layout Standard
My personal predilection is for a rather conventional laboratory experimental
approach.
Forward progress is achieved thorough a mixture of theoretical articulations
and experimental realizations.
Try to build a system that
\begin_inset Quotes eld
\end_inset
works
\begin_inset Quotes erd
\end_inset
.
Collect the data.
Analyze the data.
Try to understand
\begin_inset Quotes eld
\end_inset
what happened
\begin_inset Quotes erd
\end_inset
.
\end_layout
\begin_layout Standard
The
\begin_inset Quotes eld
\end_inset
lab
\begin_inset Quotes erd
\end_inset
here is software.
Software is commonly distinguished into imperative vs.
functional programming styles.
Off in a neglected corner is the descriptive or declarative style.
The declarative style splits computation into two pieces: a description
of
\begin_inset Quotes eld
\end_inset
what is
\begin_inset Quotes erd
\end_inset
, and a distinct component of
\begin_inset Quotes eld
\end_inset
how it interacts
\begin_inset Quotes erd
\end_inset
.
The description of
\begin_inset Quotes eld
\end_inset
what is
\begin_inset Quotes erd
\end_inset
can be understood as a description of structure, of network relationships,
of graphical vertexes and edges (or dots and arrows, for category theory/topoi.)
The
\begin_inset Quotes eld
\end_inset
how it interacts
\begin_inset Quotes erd
\end_inset
portion is a dynamical system or driver that mutates structures and relationshi
ps over time.
The driver is more–or–less structure agnostic; the driver rearranges the
descriptive/declarative program through combinatoric forces, e.g.
free energy principles, maximum entropy principles, and/or related non-equilibr
ium thermodynamical principles, including conventional ideas formulated
\emph on
a la
\emph default
Gibbs and Boltzmann, Markov/Bayesian blankets,
\emph on
etc.
\emph default
That is, the driver is presumed to accord well with generally–accepted
conventional concepts.
\end_layout
\begin_layout Standard
Accepting this split between declarative expressions of structure and drivers
of dynamical change requires a framework
\emph on
aka
\emph default
language for declarations.
The framework to be used is Atomese
\begin_inset Foot
status open
\begin_layout Plain Layout
See https://wiki.opencog.org/w/Atomese
\end_layout
\end_inset
Note that there are two distinct versions of Atomese; one that is an evolution
of
\begin_inset Quotes eld
\end_inset
classical Atomese
\begin_inset Quotes erd
\end_inset
, and another, developed from scratch, in Ben Goertzel's Hyperion system,
which is sometimes called
\begin_inset Quotes eld
\end_inset
Atomese 2.0
\begin_inset Quotes erd
\end_inset
.
For practical and technical reasons (including stability, debugability,
scalability and performance), the original formulation of Atomese will
be used.
\end_layout
\begin_layout Section*
Expected Outcomes
\end_layout
\begin_layout Standard
An experimental approach generally precludes fireworks and stymies hyperbolic
claims of grand achievements.
There's a certain rhythmic daily grind involved in constructing, running
and analyzing.
\end_layout
\begin_layout Standard
One of the simplest sensori–motor systems is that of walking through a file–syst
em, and
\begin_inset Quotes eld
\end_inset
observing
\begin_inset Quotes erd
\end_inset
what is there.
This is a very different domain than the conventional 3D abstraction of
Minecraft, virtual worlds or robotics.
It is also distinct from the 2D pixelated ARC Prize ARC-AGI-2 challenge.
Nor is it one–dimensional, in the way that streams of words/text are treated
in LLM's and transformers.
The hierarchical structure of a file system seems closer to the structure
of
\begin_inset Quotes eld
\end_inset
knowledge in the abstract
\begin_inset Quotes erd
\end_inset
or
\begin_inset Quotes eld
\end_inset
information in the abstract
\begin_inset Quotes erd
\end_inset
, and thus may provide a better experimental domain for how a knowledge–processi
ng system ingests, manipulates, excretes, reorganizes and retains structural
relationships.
(Atomese is itself hierarchical tree–structured.)
\end_layout
\begin_layout Standard
The expected object of study is a self–organized collection of structures
that exhibit navigation and movement through its environment (for example,
a file system), together with interaction with and mutation (manipulation)
of the environment.
\end_layout
\begin_layout Standard
The expected outcome is the formal (algebraic, mathematical, algorithmic)
description of such a system and a characterization of its behavior in
statistical and observational terms.
This includes scaling behaviors, measures of turbulence, observation of
avalanching behavior, movement, localization,
\emph on
etc.
\end_layout
\begin_layout Section*
Timeline
\end_layout
\begin_layout Standard
This project is already ongoing, with bits and pieces scattered across four
or five distinct github sites.
\begin_inset Foot
status open
\begin_layout Plain Layout
These include:
\end_layout
\begin_layout Itemize
https://github.com/opencog/sensory
\end_layout
\begin_layout Itemize
https://github.com/opencog/motor
\end_layout
\begin_layout Itemize
https://github.com/opencog/evidence
\end_layout
\begin_layout Plain Layout
The content at these sites is tagged
\begin_inset Quotes eld
\end_inset
version 0.1
\begin_inset Quotes erd
\end_inset
and consists primarily of design documents and a smattering of code, demos,
unit tests and examples.
\end_layout
\end_inset
Much of the effort so far has been an articulation of
\begin_inset Quotes eld
\end_inset
what the problem is
\begin_inset Quotes erd
\end_inset
, and attempts to formulate it in functional code.
The build–test–retry cycle is short, on the scale of weeks.
This is a small project; there is no plan to create a plan, with project
planning time–lines or Gantt charts.
The goal is to just
\begin_inset Quotes eld
\end_inset
move forward
\begin_inset Quotes erd
\end_inset
, as best as can be done.
\end_layout
\begin_layout Section*
Budget Justification
\end_layout
\begin_layout Standard
Hardware compute resources have already been procured, and so hardware expenditu
res of less than $10K are expected.
The nature of this research does not seem to require the use of, or interaction
with LLMs, GPT, or any of the deep–learning neural net systems; thus a
budget of zero for LLMs is anticipated.
\end_layout
\begin_layout Standard
Some of the contemplated theoretical models might run much faster if ported
to GPU's.
Perhaps the theory may indicate that the descriptive, declarative systems
resemble TensorFlow.
If so, then a contingency must be set aside for rental of cloud–compute
resources.
However, the need for this, and the nature of this need is entirely unclear.
There's a shadow of a hint at the edges; how it comes into play remains
unknown.
The current work is a low–level exploration of theory, and has not reached
the stage of being compute–hungry.
\end_layout
\begin_layout Standard
The budget is primarily that of the salary for the lead researcher.
\end_layout
\begin_layout Section*
Team and collaborators
\end_layout
\begin_layout Standard
Team of one: myself: Linas Vepstas.
The research above is sufficiently arcane, and the description of it is
sufficiently obtuse that I have not been able to attract interested parties
or collaborators.
I seem to be operating in a corner of the noosphere that gets no visitors.
So it goes.
\end_layout
\end_body
\end_document