Just completed Chapters 7, 8 & 9 of the Hidden Spring, having been looking forward to 7 when I finished 6, here. Wow, this is good stuff from Solms.
The contents / subjects …
- Free Energy Principle – Homeostasis, “self-evidencing” in self-organising systems, statistical thermodynamics, entropy and information, efficiency. Markov-blankets between self-organised layers, two-way / circular causality.
- Predictive Hierarchy – exploits these layers – more efficiency. The emergent objects (wrapped in their blankets) really are part of the functional / causal processes, without elemental reductionism. The contents of any Markov blanket are highly self-contained. (As an engineer myself, this is pure systems thinking.)
- Consciousness Arises – subjective consciousness too(!) – *BOOM* – how could it be any other way?
So much to agree with just a few meta / detail points & questions:
Efficiency as a driver? (Tim Kueper was sceptical about this in a previous exchange. The kinematic teleology is in the self-organisation, the homeostasis, free-energy efficiency is simply the mechanism. Does that help Tim?) “All the ‘quantities’ in a self-organising system that can change will change to minimise free energy”.
That “self-evidencing” is so important – counter intuitive like the directions of information flows in the first 6 chapters – most “information” about the outside world comes from inside our heads. It really does.
This is all very recent from Solms. Up to Ch6 it is mostly his career experience in “neuro-psycho-analysis“. But these chapters all arose out of seeing Karl Friston give a “Life as we know it” presentation at Wellcome / UCL in 2017, working with him on a joint paper thereafter and creating / publishing this book in 2021. Astonishing how quickly things fit into place. Even starting with an information systems bias I’ve been navigating this minefield for 20 years. (I’ll come back to the minefield / Rubicon / catch-22 / leap-of-faith element later. It never goes away.)
As well as dedicating the whole book to Panksepp, Solms credits Friston by prefixing his name to several of the effects he names here (eg “Friston’s Free Energy Principle”, Friston’s Law, etc).
Having been aligned on so much up to Ch7 he reports that Damasio parted from the mechanistic / reductionist implications of “algorithms” that flow through 7, 8 & 9 (& the Solms-Friston paper itself). It’s a common fear.
QUESTION to Mark Solms :- I wonder if Damasio might reconnect with this thesis if he sees the qualitative / categorical nature of the algorithms (p193) as opposed to long causal reductionist chains of classical objects? More heuristics than formal algorithms?
As well as Friston & Damasio – every other source in there. Markov (& Tolstoy), Shannon, Sacks, Wheeler (it from bit), Kant, Darwin, Wiener (cybernetics), W Ross-Ashby(!), Gibbs, Helmholtz (but not Boltzmann or Mach), Freud (but not Maslow), Varela. Wonderful stuff.
Ergodicity – it’s been my favourite concept since 2017 – he uses ergodic and non-ergodic a couple of times in a long Friston quote (p162/3) without any definition or clue to intended meaning. (2017 a coincidence, no parallel connections). Mentions it again in another Friston quote (p170) with “An ergodic system occupies limited states”. And then relies on this to conclude that all (biological) self-organising systems:
- are ergodic
- have a Markov blanket
- exhibit active inference
- are self-preservative
Apart from questioning whether that is the whole intended meaning of ergodic here (?) the active inference is absolutely key. Internal models / algorithms are constantly checked against external sense data, to improve the ongoing model (as well as fix the current homeostatic error)! (This is taking me back to 1989-ish.)
(Whether that’s all there is to ergodicity may not matter here, the key thing is it supports the limited qualitative / categorical options available to our sub-system rather than quantitative resources from across an entire population of all fine-grained options over the map.)
Anyway, joining up these free-energy / homeostasis arguments with the previous six chapters:
“The mistake most cognitive scientists make is that they assume most incoming data is ‘exteroceptive’. They forget that expectation errors (sensory inputs) that matter most to us come from within.
These signals generate ‘affects’ not perceptions. As Freud said, the forebrain is a ‘sympathetic ganglion’. Confusion on this score is the perennial price most cog-sci’s pay for adopting the cortical fallacy. Consciousness is endogenously generated, all of it. Consciousness at source is affect.”
“Affective valence – our feelings about what is biologically ‘good’ and ‘bad’ for us – guides us in unpredicted situations. We concluded that this way of feeling our way through life’s problems using ‘voluntary’ behaviour is the biological function of consciousness.”
That leaves me with one clarifying thought
QUESTION to Mark Solms: – those feelings of affective valence are about what is good or bad for us biologically including psychologically / mentally? ie Useful work for free energy includes thinking (feeling) about our thinking as well as as thinking (feeling) about our acting in the world?
There’s a good deal of rehabilitation and building on Freud – beyond the usual caricature of dreams and parental influence (where Solms started in fact) – in areas where he really was ahead of his time on how brains / minds work. Solms (and Friston) developed formal set of equations (using Freudian notation) to describe the algorithmic information behaviour between the various brain states in play, neatly summarised in “Fig 17” where that “mid-brain decision triangle” behaviour is laid bare. It does look bonkers to “reduce” mind to such a simple graphic information flow diagram of a few equations(!) no less – see eg Damasio’s reaction earlier – but with the right perspective on the affective categorical variables in play, and accepting the infinite sub-divisibility of the reality of sub-systems within any Markov-blanketed system we choose – it really is credible and convincing (it’s the delegated / permissive supervisory control system I described at the end of the previous post.).
(c) Mark Solms.
(I need to create a slide version of that with more labelling in the graphic itself.)
THE PROBLEM, as Solms elaborates in the closing pages of Ch9, is what he calls inviting scientific sceptics to cross a Rubicon. As noted earlier this tricky step never goes away from this debate, philosophical or scientific. I’ve been calling in Catch-22 for two decades, but whatever language you choose:
“[Mind is primarily affective, felt subjectively.] To rule the subjective perspective out is to exclude from science the most essential feature of the mind.”
Solms (and I) invite you to take that perspective of subjective self-hood, the one you already “have in mind” into your scientific considerations of mind.
“I am asking you to replace the third-person objective perspective we have taken so far on the dynamics [of the neuro-science] with a first-person one: with the subjective perspective of the self-evidencing system itself. I am asking you to adopt the system’s point of view, to empathise with it.”
DO NOT PASS GO.
(A little under 100 pages to go.)
In that year-old post where I first referred to Mark Solms – as something that looked important even if I didn’t find time to read – I was clearing out lots of bookmarks that were choking off any sense of making progress. One of those book marks was to this piece by Karl Friston et al on the switch from Cartesian Dualism to “Markovian Monism” – and it includes variations on that Fig 17 above. It’s all connected!
Also had a response overnight from Solms on the two questions above:
Excellent. Thanks for your response.
Having now completed it, I find my second question is thoroughly answered in the next chapter. Will compile a fuller review.
— What, Why & How do we know? (@psybertron) March 7, 2022
That later Damasio reference:
Man, K. and Damasio, A. (2019) “Homeostasis and Soft Robotics in the Design of Feeling Machines”. In Nature Machine Intelligence, I: 446-52, doi.org/10.1038/s42256-019-0103-7
If he’s OK with “feeling machines” it sounds like he’s lost his aversion to the risk of reductionism in mechanistic algorithms 🙂 Yay!
(Earlier 2010 Friston reference in there too – unfortunately the Nature Reviews – Neuroscience papers are not free access.)
On to a fuller review of Solms.
2 thoughts on “Before Returning to the Cortex”
Nice post Ian (and nice blog by the way).
I have no argument at all with efficient learning, or with the brain seeking the least difference between its internal model and the perceptions it is modeling. Looks like good solid mechanism to me.
My skepticism would be reserved for an attempt to take that internal teleology and move it outside of the brain, to the goal of life itself. In both cases, efficiency is a key for a self-organizing system, but the energy considerations are inverted. The brain is trying to efficiently minimize the free energy of a gradient (which in this case represents surprise). The brain here is like a tornado, trying to reduce a gradient as fast as possible. Life, on the other hand, is trying to efficiently sustain itself at a higher energy level (trying to delay the dissipation of sunlight). The goal of the brain is to minimize error as quickly as possible, but that goal is in service to the goal of life, which is to persevere in being.
I do have some questions about the boundary between sacred nature and objective science. I was going to wait for your next post on left hemisphere/right hemisphere before asking about that.
Excellent, on the free-energy / 2nd law / efficiency & “teleology” – we’re getting there.
Brain like a tornado – I may steal that. All-scale compressible flow analogies really seem to work.
Ah yes – “sacred naturalism” or “natural theology” is becoming a key topic.
I need to complete my write-up of Solms first (completed the read last night).
The “consciousness as affect” is kind of part of it – naturalism beyond objective science.
Thanks for the engagement (Solms himself responded to my questions on Twitter – so looking good.)