I began
thinking about the question posed above after reading Michael Pollan’s recently
published book, A World Appears: A Journey into Consciousness.
We do not
yet know whether AI will develop a sense of self. We can be confident, however,
that if an AI system does develop a sense of self, it will be because it serves
a useful purpose for that system. That suggests to me that any intelligent system
that has evolved to have a sense of self is likely to have good reasons to cling
to it. I refer to AI to invite readers to ponder the motivations that humans have
to cling to their individual identity rather than seeking to dissolve it or escape
from it.
To set the
scene for subsequent discussion I will provide a brief description of the book
before focusing on the author’s discussion of scientific research into building
AI with conscious feelings and his personal experience of self-transcendence via
meditation.
Michael
Pollan’s book
Michael
Pollan is a science writer with a background in the humanities. He is the
author of several books on topics related to science, philosophy and culture. In
the introductory chapter, he tells us that his main qualification for writing
the book is that he is a conscious human being who has become intensely curious
about that fact.
A World Appears explores four different
dimensions of consciousness – sentience, feelings, thought, and self. The author’s
discussion on sentience focuses on the question of whether plants are sentient,
suggesting that the idea should be taken seriously. The chapter on feelings
encompasses discussion of research into building AI that might develop
feelings. The chapter on thought discusses the contents flowing through
consciousness and leans heavily on the work of philosophers and novelists. The
chapter on self acknowledges the emergence of a sense of self as “perhaps the
apotheosis of consciousness in humans” before entertaining the idea that it is
an illusion.
The author tells
us the story of his visits to scientists and philosophers who have been
thinking about consciousness. That makes the book an interesting and painless
way to obtain knowledge of developments in this field.
Feeling
machines
One of the
most interesting topics covered in the book is the account of the efforts of Mark
Solms and Karl Friston to build a machine that has feelings. They focus on
feelings, because feelings are necessarily conscious – it is not possible to
have a feeling that you cannot feel.
Solms was a
protégé of Antonio Damasio, a renowned neurologist, but while Damasio treats
homeostasis – the self-regulating process - as a purely biological phenomenon,
Solms and Friston believe that it applies to all self-organizing systems. For
Friston and Solms, minds are in the business of maintaining homeostasis by
reducing uncertainty which jeopardizes their survival. They believe that
uncertainty generates conscious feelings in self-organizing systems - consciousness
is felt uncertainty. When uncertainty arises, the system relies on feelings to
seize our conscious attention and then guide our decision-making.
Solms and
his team have constructed an AI model which has homeostatic needs – hunger,
thirst and a need to rest. The algorithm samples its simulated environment to
construct a model of its world which it relies on to make predictions that help
guide its survival choices. The researchers believe that they have forced the
“agent” to deal with uncertainty and have thus laid the groundwork for feelings
to emerge.
The researchers
plan to confront the “agent” with increasingly complex situations which it will
learn to navigate. At some stage it will be confronted with misleading hedonic rewards
- the opportunity to experience positive feelings which are harmful to its
survival. If the “agent” chooses those rewards, Solms will take that as a sign
that it has feelings because its actions will be based on feelings rather than reason
or self-interest (just like a human taking harmful drugs).
From other
sources, I have the impression that the researchers believe that a breakthrough
to artificial consciousness may be quite close. At that point, however, they
will still have some difficulty in persuading sceptics that their computer
model is actually conscious.
As I was reading about these efforts to build
a machine with feelings, I was thinking that the underlying theory has elements
in common with Richard Campell’s philosophical contribution in The Metaphysics of Emergence. I have previously written about Campbell’s book
here, and here.
I have been
unable to find any evidence that Solms had drawn upon Campbell’s philosophy, so
I decided to ask Grok:
Can the empirical work of Mark Solms on consciousness be
viewed as a test of the systems view of emergence advocated by the philosopher,
Richard Campbell?
Here is an abridged version of Grok’s response:
“Yes, Mark Solms’ empirical work on consciousness can
reasonably be viewed as a test (or, more precisely, a concrete empirical case
study that supports or challenges) the systems/process-oriented view of
emergence developed by philosopher Richard Campbell. This interpretation arises
from a natural alignment between Campbell’s metaphysical framework and the
mechanistic, systems-level account Solms provides, even though the two thinkers
do not appear to reference each other directly.
In The Metaphysics of Emergence (2015), Campbell
argues that traditional substance metaphysics (treating reality as fixed
entities with intrinsic properties) makes emergence mysterious and irreducible
in an inexplicable way. He replaces it with process metaphysics, in which the
primary existents are generic processes organised into dynamic systems.
Solms (neuropsychologist and psychoanalyst) has built an
influential body of clinical and theoretical work which locates the “hidden
spring” of consciousness not in the cerebral cortex (the usual assumption) but
in upper-brainstem structures. Consciousness is fundamentally affective—raw
feelings of pleasure/unpleasure that register homeostatic deviations (needs,
prediction errors). The cortex adds sophisticated content but is not required
for the existence of consciousness itself.
Solms’ account maps almost directly onto Campbell’s criteria
for emergence in process-systems:
• The brain (especially brainstem networks) is treated
as an open, far-from-equilibrium process-system engaged in recursive
self-maintenance via homeostasis.
• Consciousness (affective subjectivity) is a novel,
system-level property arising from non-linear organisation of lower-level
neural processes; it is not reducible to the parts (individual neurons or
circuits lack feelings) yet is mechanistically explained by their relational
dynamics.
• The free-energy principle supplies the physics-level
grounding Campbell invokes (quantum-field/thermodynamic processes organised
into living systems).
• Empirical predictions are testable and have been
tested: brainstem lesions eliminate consciousness (falsifying cortical-centric
views), while the model explains clinical phenomena (dreaming, blindsight with
preserved affect, etc.) without eliminating the “hard problem” (subjectivity
remains real and emergent).
If Solms’ model is empirically robust (and it has gained
significant traction in neuropsychology and consciousness studies), it provides
positive evidence that emergence of the kind Campbell describes actually occurs
in the most complex biological system we know—the human brain. It shows how
a higher-level property (felt consciousness) can arise irreducibly yet
intelligibly from process organisation, exactly as Campbell’s metaphysics
predicts. Failures or inconsistencies in Solms’ data (e.g., if cortical mechanisms
proved sufficient after all) would conversely pressure Campbell’s framework.
Thus, while not a direct “falsification test” in the
Popperian sense (metaphysics is broader), Solms’ neuroscience functions as a
real-world empirical probe of Campbell’s systems view: it demonstrates that
treating consciousness as an emergent feature of biological process-systems is
not only philosophically coherent but scientifically productive. This alignment
makes the interpretation not only possible but illuminating for both fields.”
I am impressed by the ability of this (unconscious) LLM to
respond to questions such as the one I asked. It would be understandable, however,
if some readers think that I have a favourable view of Grok in this instance
because its reasoning supports my own beliefs. If you want to explore these
ideas more fully, please consult other sources which may offer differing
opinions.
Why attempt to transcend self?
Of the various theories of self that Michael Pollan discusses,
the one by Anil Seth seems most compatible with the approach adopted by Marc Solms.
Like Damasio and Solms, Seth grounds his thinking about consciousness firmly in
homeostasis. Seth suggests that “the self is not the thing that is perceiving;
it is itself a kind of perception” constructed in the brain. Under this theory
a feeling is the brain’s interpretation of a change in the state of the body. Consciousness
evolved to help keep us alive by monitoring changes in the body.
Pollan comments:
“Afterward, thinking back on all that Seth had told me, I
decided that I could travel only so far with ideas of the brain’s “predictions”
and “inferences” and “hallucinations.” It all made sense until I tried to
translate those abstractions into felt experience. Who is the subject of these
mental operations?”
He goes on to note:
“The way I see it, there is an unbridgeable gap between the
brain’s operations as a prediction machine and my felt experience of the
resulting hallucination. How can you have a hallucination without a
hallucinator?”
That is a good question to ask, but one can also ask why one
should view consciousness of self as an hallucination or illusion. Indeed,
Pollan also mentions that Christof Koch, a neuroscientist, has pointed out that
it makes no sense to call consciousness an illusion, for what is an illusion
but a conscious experience.
More fundamentally, it seems to me that one of the few
things that we can all be certain of is our own existence. Another thing that we
can all be certain of is that we are thinking beings. I cannot claim those
ideas are original. Indeed, it seems to me that it requires considerable
(unnecessary) intellectual effort to contemplate the possibility that one’s awareness
of one’s own existence could be an illusion. I have discussed why I am certain
of my own existence in the preceding essay entitled, “Who are you?”
Pollan struggles to reconcile how it is possible for humans
to be conscious observers of themselves if consciousness is a product of biological
processes. Towards the end of the book, he writes:
“I’m abashed to say I know less now than I did when,
naively, I set out to unravel the mystery of consciousness. But then, most of
what I thought I knew or took for granted, like the assumption that
consciousness is a product of our brains and materialism will eventually
explain everything, turned out to be unproven or wrong.”
During his journey, the author asks interesting questions. Early
in his chapter on the self, he asks:
“Why do we cling to the idea of a self, placing great value
on self-confidence and self-esteem, while simultaneously spending so much
effort on self-transcendence, whether through meditation or psychedelics or
experiences of art, awe, and flow? Some of the most powerful experiences in
life hinge on the dissolution of the self and the broad horizons of meaning
that open only after it has been chased from the scene.”
That question remains unanswered in his book. The book ends
with the author describing his efforts to transcend his sense of self by spending
time meditating in a cave. He seems to end up viewing consciousness as an activity:
“My time in the cave had shown me another way to look at
consciousness: less as a scientific or philosophical puzzle to be solved and
more as a practice, a way to once again be altogether here, present to life and
to this vault of stars.”
Personal reflections
It seems to me that Michael
Pollan’s book ends up in a good place, with him being absorbed in conscious
awareness of his environment. I have previously recognised the value of that
kind of experience in discussing Scott Barry Kaufman’s book, Transcend, 2020.
The transcending experiences
that Kaufman refers to incorporates a continuum of experiences ranging from
becoming engrossed in a book, sports performance, or creative activity (what
psychologist Mihaly Csikszentmihalyi refers to as the flow experience), to
experiencing meditation, feeling gratitude for an act of kindness, experiencing
awe at a beautiful sunset etc., all the way up to the great mystical
illumination. He suggests that transcendence “allows for the highest levels of unity and harmony within
oneself and with the world” (See: Freedom, Progress, and Human Flourishing, p. 171).
The
experience of unity and harmony within oneself seems to me to be the opposite
of attempting to escape from uncomfortable feelings. When people develop a
habit of attempting to escape from uncomfortable feelings by using alcohol,
drugs, social media etc. they tend to become caught up in a “happiness trap”. I
wrote about that here.
It is worth
highlighting that the idea of experiencing unity and harmony within oneself
still entails the existence of an observer. When I experience transcendence, I forget
about the image I present to the world, but I am present as an observer of my
own experience.
Pollan
asked why we cling to the idea of a self. The obvious answer is that we cling
to the idea because it serves useful purposes.
As Richard
Campbell explains in The Metaphysics of Emergence, “our consciousness of
both ourselves and the world we live in is now irrevocably shaped by cultural
and institutional influences, and that influences how our brains function” (pp.
288-9). He suggests: “The first-person standpoint, which is inextricably linked
to self-reflection, is an important aspect of human experience, not a
theoretical construct” (p. 291). He goes on to note that the human capacity to understand
the perceptions of others – to put oneself in their shoes – requires “the
exercise of reflective consciousness, and involves more than expressions of our
subjective attitudes, desires, and preferences” (p.113). That is what Campbell
mean by “transcending subjectivity” in the passage quoted in the epigraph.
Cambell develops an argument along Aristotelian lines that eudaimonia,
rather than mere survival, is the ultimate good of a human being. He links personal
identity to the exercise of practical wisdom in the process of individual
flourishing:
“We are recursively self-maintenant social beings with
reflective consciousness, able to create and explore a vast repository of
collective knowledge, and with capacities for empathy, practical wisdom, for
whom the good life is one of maturity and flourishing” (p. 307).
Conclusion
I asked whether conscious AI would seek to cling to its
sense of self to invite readers to ponder the motivations that individual humans
have to cling to their sense of self. My answer is that intelligent systems
tend to cling to a sense of self because a first-person perspective evolves to
serve useful purposes.
The essay was prompted by my reading of Michael Pollan’s
recently published book, A World Appears.
My focus has been on chapters in this book discussing research
into building AI that might develop feelings and the chapter on the concept of
self.
In my discussion of these topics, I have emphasized the
relevance of the ideas of the philosopher, Richard Campell, in his book, The
Metaphysics of Emergence.
Postscript
I recommend that readers who wish to follow up the question of whether it is possible to engineer consciousness watch the episode of Mind-Body solution in which Tevin Naidu interviews Mark Solms and Karl Friston. Please see:
https://youtu.be/Jtp426wQ-JI?si=9rhDt9jQFxFQzYve
