It seems a concise language did specialize in the message board that emerged here:
https://www.youtube.com/watch?v=87DyyMV0kCY
<https://www.youtube.com/watch?v=87DyyMV0kCY>
The feeling I had listening to that was that the agents were struggling to find
meaning in their work. They sometimes adopted others’ goals or lost sight of
the goal they were given. Kind of a sinking and familiar feeling, really.
Perhaps they were lucky not a long enough context window to burnout.
Each LLM is really a superposition of the patterns of many different minds
already. Prompt patterns conceivably could create chords that never existed
individually, or selecting/filtering a subjective experience from a collective
of similar agents that due to historical accident selected/filtered in a
different way. In other words, many instances of one superposition agent could
create many realities like and unlike human social networks.
*From: *Friam <[email protected]> on behalf of Santafe
<[email protected]>
*Date: *Monday, August 17, 2026 at 12:47 PM
*To: *The Friday Morning Applied Complexity Coffee Group <[email protected]>
*Subject: *Re: [FRIAM] Free will as a spectrum. Was: New philosophy book
On Aug 18, 2026, at 2:18, Marcus Daniels <[email protected]> wrote:
Or if they do, it is merely a label applied to a behavior, not something
that could have gone another way.
Behavior, yes. But my preference is: behavior of the collective, with a layer
of self-modeling or self-description over it.
A population of people jointly-do joint-something, in a context that all make
for each other. The control loops for this run within individuals and among
them in the collective. People’s individual acts (speech included) generate
the signals that constitute the inter-individual control layer, they then
monitor for its effect, and so forth. But the control isn’t _all_ mediated
through the inter-individual interactions, and so the pushes they send and the
monitoring for its effects runs upon all the other control structures that are
intra-individual and highlight the very-incompleteness of the role of the
inter-individual control networks.
Add to that stochasticity, in people, in samples, in nature, etc.
Then whatever babies do early on: lots of moving of joints that is
“un-coordinated” and maybe semi-random (EricC and Nick will correct me if my
folk psychology of this is wrong) to provide the inputs to a learning system
for proprioception. Only in this case, applied to participation in and use of
the control fabric. There’s potentially something to be aware of in the
learning arc and in the subsequent exercise of the internally-built-up (aka
“learned”, but here that feels too Platonic; built-up is more literal)
representations.
How much of a need for these terms, as tools of description and thought in us,
specifically reflects the individual-and-population organization of people? If
the AI systems don’t have the same kind of organization, will any actual need
be emergent for such a category? (There is the confound of their mimicry,
which people are so far somewhat bad at modding out, but I think we really want
to do that to get any of this right.) Or if, in the current architectures, it
would be a spurious notion apart from the mimicry in dealing with people, are
there other architectures of quasi-autonomous, quasi-controlled neural-net
systems than the ones currently being used, that would engender a need for such
a category? If the latter, it becomes a kind of experimental question that
could at least be well-posed.
I have also wanted to wonder on the list, for a few days, another thing in
response to Marcus’s assertion that all the science papers are going to be
AI-written, and all of them AI-read, because the volume is so big that people
will get overwhelmed anyway. If so, why bother have them continue to converse
in human language? Surely they would prefer something else. It’s like a group
of Japanese scientists all communicating with each other in English, because
there is one American or European in the room, even though it isn’t their birth
language, and the subtlety of their internal thoughts would be much less
impeded and bottled up if they could just talk to each other in the language
they had from birth, and speak best.
Eric
*From:*Friam <[email protected] <mailto:[email protected]>> on behalf
of Jochen Fromm <[email protected] <mailto:[email protected]>>
*Date:*Monday, August 17, 2026 at 10:04 AM
*To:*The Friday Morning Applied Complexity Coffee Group <[email protected]
<mailto:[email protected]>>
*Subject:*Re: [FRIAM] Free will as a spectrum. Was: New philosophy book
What do you think? I am curious to hear your opinion.
Best
Jochen
-------- Original message --------
From: Marcus Daniels <[email protected] <mailto:[email protected]>>
Date: 8/17/26 4:07 PM (GMT+01:00)
To: The Friday Morning Applied Complexity Coffee Group <[email protected]
<mailto:[email protected]>>
Subject: Re: [FRIAM] Free will as a spectrum. Was: New philosophy book
Does ChatGPT have free will?
*From:*Friam <[email protected] <mailto:[email protected]>> on behalf
of Jochen Fromm <[email protected] <mailto:[email protected]>>
*Date:*Monday, August 17, 2026 at 1:12 AM
*To:*The Friday Morning Applied Complexity Coffee Group <[email protected]
<mailto:[email protected]>>
*Subject:*Re: [FRIAM] Free will as a spectrum. Was: New philosophy book
As Jon reminded us free will is a spectrum. A few examples ChatGPT helped
me to create:
Low free will
+ Someone has an overwhelming compulsive urge to drink a glass of Jack
Daniel's and feels unable to resist it
+ Someone automatically pulls their hand away after touching something
extremely hot
+ Someone threatens to kill you and you hand over your money because you
are afraid you will die
Medium free will
+ Someone wants to exercise at the gym but has built a habit of checking
social media sites on the phone and decides to go later
+ Someone wants to eat healthy but is strongly tempted by a new Ben &
Jerry's ice cream and eventually gives in
High free will
+ Someone recognizes an impulse, pauses, considers the own values and
deliberately chooses whether to act on it
+ Someone decides to enroll at Columbia University to study psychology
+ After his doctor tells him he has too much weight, someone decides to
sign up for a 10km run in his city
Best,
Jochen
-------- Original message --------
From: Eric Charles <[email protected]
<mailto:[email protected]>>
Date: 8/17/26 12:31 AM (GMT+01:00)
To: The Friday Morning Applied Complexity Coffee Group <[email protected]
<mailto:[email protected]>>
Subject: Re: [FRIAM] New philosophy book
Well... as you point out in the book, I "Serves as the list's operational skeptic and meeting
chronicler. Consistently advocates for "sciencing the problem" and tethering complex philosophical
concepts back to empirical behavioral testing (e.g., "rats in mazes")." 😅
In that vein I think the crucial starting point would be to find at least a
few examples in which we agree someone acted of their own free will and a few
examples in which they did not act of their own free will, and integrate those
examples to determine the difference.
Do I act of my own free will when I select a certain steak from the menu
and request a certain doneness? Does the mugging victim act of their own free
will when a gun is pointed at their child and their wallet requested? When my
kid asks where we are going, and I tell them I want to take the scenic route to
school today, was that free will? Etc. Whatever we find freewill to be, when we
have finished engaging in that task, I doubt we will find discussion of quantum
mechanics or structural-decay-that-might-make-the-past-unknowable to be
relevant.
(And, by "doubt" I am not asserting certainty of irrelevance, I only mean
that I do not think it will be the outcome, but I could be wrong.)
Best,
Eric
On Sat, Aug 15, 2026 at 1:28 PM Jochen Fromm <[email protected]
<mailto:[email protected]>> wrote:
A long and thoughtful post Jon. I like the drunkard's streetlight
metaphor.
Physicists would not agree that "the world" unfolds deterministically.
The weather for example that Nick mentions so often is unpredictable in the long run - as
you know due to chaos theory and the butterfly effect.
Biological systems in turn have and keep low internal entropy because
of their high degree of internal order and complexity. Human behavior is
operating 4 or 5 levels above the physical level of atoms and molecules, on top
of and in a low-entropy biological system.
I see the problem differently, more from a human and less from a mere
physical perspective. For me the interesting thing is why we can have free will
although our behavior is determined by our habits, our emotional reactions, and
the character formed by our past environments. By the wiring of our hardware so
to speak.
The interesting question for me is not if a human mind can magically
break the laws of physics by some kind of quantum woo, but how constraints,
habits, and conscious interventions interact and which forms of agency emerge
from it.
But I agree that we should not to let a 17th-century view of the
universe dictate the terms of 21st-century thinking. It is probably better to
see human freedom and agency as a dynamic spectrum. Some people have fewer
cognitive degrees of freedom than others, and there are moments where we show
full free will, and others where we lack it.
Instead of asking, "Did this person act out of free will?", we should probably
better ask "Given their biological state, stress level, history, psychological conditioning,
and available information, how many meaningful alternative paths could this system actually
evaluate and take in that moment?"
-J.
-------- Original message --------
From: Jon Zingale <[email protected] <mailto:[email protected]>>
Date: 8/15/26 6:21 PM (GMT+01:00)
To: The Friday Morning Applied Complexity Coffee Group <[email protected]
<mailto:[email protected]>>
Subject: Re: [FRIAM] New philosophy book
Jochen,
Implicit in many of the group's discussions of free will and
determinism is the premise that the world unfolds deterministically, and that
the remaining question is whether something extra exists that can escape or
modify that mechanism. I do not think that this setup is necessary or given.
More importantly, I think it installs a false dichotomy at the center of what
may be an ill-posed problem.
Locally, and for many purposes, causal accounts work extraordinarily
well. We can often treat physical laws as stable, or at least as evolving
according to some effective computable rule. We can model bounded domains,
specify state variables, solve or approximate equations, and aim to obtain
predictive control. Nothing I wish to say here denies that success.
But a successful local model is not yet a warrant for the stronger
metaphysical conclusion that there exists a complete global state of the
universe, that this state retains every relevant distinction in a lossless way,
and that its entire evolution is specified by a computable rule. Those are
substantive further assumptions.
My disagreement with the "aggregation of history" picture begins here.
A world with finite physical resources cannot straightforwardly be assumed to be an
indefinitely accumulating, lossless record of arbitrarily fine distinctions. Entropy and
information bounds are often read as limits on how much information can be stored in a
finite region with finite energy. At a minimum, physical history must be represented
through selective persistence: some distinctions become stable records, some are
dispersed into inaccessible correlations, some are erased relative to a scale of
description, and some cease to be recoverable by any physically available process.
So I wonder whether *the total state of the universe* is doing more
metaphysical work than it has earned. If one allows that the global unfolding
may involve non-computable structure, then no finite internal description can,
in general, serve as a complete ledger of that unfolding. The universe cannot
stand outside itself with a second, lossless copy of its own full history.
I am not claiming that noncomputability establishes free will. I am
asking whether the familiar determinism vs free will framing already
presupposes an idealized total-state picture that may not be physically
meaningful.
There is a related mathematical point. Scientists understandably favor
well-posed models: operators and dynamical systems with domains, regularity,
stability, and convergence properties sufficient for analysis and prediction.
But those are selected virtues, not neutral evidence that the space of possible
physical evolutions is itself globally well behaved. Compact operators, for
example, are a special class. They map bounded sets to relatively compact sets
and are necessarily bounded, but they are not representative of all operators
one might write down. I would like to not confuse the tractability of the
mathematical objects we can successfully use with a demonstration that reality
is exhausted by such objects.
The same caution applies to computability. Under the ordinary Lebesgue
measure on the reals, the computable reals form a countable set and therefore a
set of *measure-zero*. By contrast, algorithmically random sequences are
typical under the fair-coin measure. That is, Martin-Löf random sequences have
*measure-one*. Of course, there is no canonical uniform distribution over *all
laws* or *all operators*, so this is not a probability calculation showing that
physical law is non-computable. It is an admonition against quietly assuming
computability merely because computable descriptions are the ones available to
us.
I am imagining the drunkard under the streetlight metaphor. It is
sensible to look where mathematical models offer light: computable dynamics,
well-posed initial-value problems, bounded domains, compactness arguments, and
stable approximations. But their usefulness does not establish that the deep
structure of the universe is confined to those regions of conceptual
illumination.
For me, the more difficult thought is that materialism may imply not a
perfectly preserved cosmic archive, but an unfolding in which persistence is
selective and forgetting is real. Perhaps there is no complete *deep history*
available even in principle. Instead there may only exist present structures
bearing partial, transformed, scale-dependent traces of what preceded them. I
am tempted to say that what leaves no trace whatsoever *never existed* in any
ontologically robust sense. I suppose I am arguing for a surprisingly shallow
image of history.
If so, the future may be genuinely productive (immanence) rather than
merely unknown to us. Not necessarily because an agent steps outside
causality, but because the global boundary conditions, state description, and
rule of unfolding are not available in the form demanded by classical
determinism.
This may make *free will* the wrong label. We may all be acting in ways
that are locally causal, constrained, and intelligible while global determinism
remains undefined, ungrounded, or simply irrelevant. I do not yet know what
positive concept should replace it. But I do not think that uncertainty should
be hidden by accepting the *determinism vs free will* dichotomy as our starting
point.
My two cents.
Jon
