A Software Engineer’s Guide to Cybernetics | Hacker News Reader

A Software Engineer’s Guide to Cybernetics

medium.com

179 pointsmbellotti59 comments

I wrote my PhD on the philosophy of Norbert Wiener, namely, 'cybernetics'.

I think what people haven't realised is that for Wiener - though not necessarily the successive generations of Cyberneticians per se - cybernetics names not only a transdisciplinary field of study, not only a primitive form of control engineering, but a philosophical system that contends directly with GW Leibniz's monadology. Wiener considered it to be a kind of monster, a 'golem', since whereas the monadology posits universal beneficence throughout nature and humankind, Wiener's cybernetics is, when applied to human societies, structurally, infinitely, conflictual. Moreover it points towards awesome consolidations of power.

Generations of readers of cybernetics have completely missed or disregarded this aspect of his thought, conveniently mistaking a system of regulation in nature for a system of regulation in artificial human societies. For Wiener there is no self-regulation in the body politic, societies are fundamentally anti-homeostatic, and the techniques that his generation had bequeathed to the world would only exacerbate its positive feedback loop.

This is why, for me, Wiener is still so compelling, despite the unfathomable equations, the desperate arrogance, the stuffy air of cigars. He was the first to see the world that Snowden brought to the public's attention. And it's all there, systematically in 1948 (and even, sort of, earlier).

Please say more.

I would LOVE to hear how Wiener might interpret today's world.

As a total noob, I'm often delighted by the insights of earlier commentators. The people who lived thru great transformations, so can remember the before times, allowing them to see the now more clearly.

Like your lament about Wiener's ignored analysis and insights, rereading McLuhan was very fruitful. His concept of autoamputation seems to have been completely ignored. (Over stimulation leads to shutting down that sensory input.) But it's such a great way of describing the impact of modern rhetoric.

Sometimes I wonder if it's like Steve Jobs' visit to Xerox PARC. Jobs admits they (Apple) were so captivated by GUIs they totally missed other breakthrough concepts like object-oriented.

Read his book, God and Golem. The essence is all there including most of his projections for the future.

Besides predicting to a T multiple aspects of the world we're now experiencing, this book is full of insinuations about several conditions yet to arise (or maybe risen but yet to be discovered!). He doesn't explicitly lay them out, but leaves you to identify and connect the dots. This occultation of ideas is somewhat common in early cybernetics works.

Needless to say, the vision of Wiener will be seen by many who understand it as supremely negative.

An alternative is explored in 'Steps to an Ecology of Mind' by Bateson.

Thanks for the cite. Here is the pdf simson.net/ref/1963/God_And_Golem_Inc.pdf

I suspect (not having yet read the above beyond the preface) the root issue is, as usual, the nature of consciousness. My own views on that subject are fringe per du jour orthodoxy which states that 'mind arises from structure'. Weiner seems to take it for granted that an algorithm "learned" something. Does an algorithm even have a "self" that would "learn"? I personally think we're being careless with word usage.

Once you contest the implicit equivalence made between 'form' and 'meaning' which occurs when one asserts that a mechanical/causal device has "learned" something -- all we have observed is form in -> form out -- then the 2 remaining "points in cybernetics" are "[machine] self-reproduction", and, "man-machine coordination".

These last two are difficult topics -only- if generously attribute "consciousness" to machines that auto-extend their state transition map. Otherwise, they are interesting but not cosmic questions. And further, one can still entertain a 'Universal Mind' and speculations such as Leibniz's monadology, happily coexiting with a science of cybernetics.

Would be interested to read that if you have it available online!

I was really disappointed when I first learned about Cybernetics, to read that it had fallen out of favor, breaking into a thousand discrete disciplines.

"Cognitive Science" seems to be the closest modern inter-disciplinary topic. I suppose it comes down to the reality that the practical applications of a true "cybernetics" discipline would be so narrow that it would be a tiny field.

I'd love to read the dissertation, can you share it?

Cybernetics also reintroduced teleology into the sciences, which is philosophically deeply controversial.

I also like the fact that he describes this teleology as reducing local entropy [1].

I presented my tenure midterm today on "AI for Wellbeing" which I based on Cybernetics [2].

[1] "while the universe as a whole, if indeed there is a whole universe, tends to run down, there are local enclaves whose direction seems opposed to that of the universe at large and in which there is a limited and temporary tendency for organization to increase. Life finds its home in some of these enclaves. It is with this point of view at its core that the new science of Cybernetics began its development."

[2] https://docs.google.com/presentation/d/1_hDjT06C6WyITgtbdPVj...

You should also look into our IEEE standard 7010 "Recommended Practice for Assessing the Impact of Autonomous and Intelligent Systems on Human Well-Being"

I was one of the authors and it might have some valuable concepts you could use.

https://standards.ieee.org/standard/7010-2020.html

Wow, thank you so much for sharing. This is, ah, quite relevant!

Would also be interested in learning more about this. Was this work recent? More generally, is there on-going research within the field of cybernetics?

curious what you think of other social systems theory folks like talcott parsons and luhmann

The first trained multilayer deep neural network was presented by the Soviets in 1970 in a paper titled “Heuristic Self-Organization in Problems of Engineering Cybernetics“ http://www.gmdh.net/articles/history/heuristic.pdf

I feel it's some kind of unwritten law that every time you think you've solved something, there's some Soviet engineer who already did it a few decades ago

I imagine the solution to every problem can be found lying on a shelf, untranslated in a research lab in the middle of nowhere in Russia

They also developed TRIZ, a Russian algorithm for inventing new systems, https://en.wikipedia.org/wiki/TRIZ

> a theory which defines generalisable patterns in the nature of inventive solutions and the distinguishing characteristics of the problems that these inventions have overcome

It's been used by Samsung and other orgs.

https://upload.wikimedia.org/wikipedia/commons/b/bf/40_princ...

Looking at that cheat-sheet, it's interesting how you can recognize elements of one's own process. I bet almost everyone can check off a few of those boxes as "I do this all the time".

Also interesting to note there are no interaction principles. "19 Pulsed Action Periodic Action" can be mapped in software systems to PUSH+Timer, but that's it. TRIZ apparently lacks the vocabulary to explore the space of interacting components [as first class "principles"].

If you ever really need to invalidate a patent, a good way is to hire a Russian-speaking agent to sift through old engineering journals and libraries. A lot of prior art turns up that nobody knows about.

Like the running gag in Star Trek TOS where Chekov would invariably assert that "Russians invented it first". Only, for real this time.

For all its faults and imperial tendencies and repression of its minority peoples, the USSR was more than just Russian. It had many peoples who all contributed to discoveries, but "Russia" today has inherited the "creds" to what the USSR accomplished, both good and bad.

(For a tiny glimpse of the USSR "feel", watch the BBC movie "Afghan experience" where some USSR soldiers from other republics than Russia proper speak.)

Especially interesting with all of the talk about the unique levels of innovation that capitalism brings that supposedly justify all of the other stuff.

Research != innovation. Innovation is more about practical implementation of novel ideas (incl. ones coming from research) - something at which soviets weren't very good at.

Could you give any examples

Innovation happens mostly in a bottom-up fashion, either on a big scale, the way it's done in Silicon Valley, or in smaller scale, by for example just having thousands of independent grocery stores in a given city, each free to experiment with the way they conduct business.

In contrast, in soviet states, everything was centrally organized and ran, which meant zero freedom to innovate at the ground level. This resulted in incredible inefficiency, as better ways of doing thing couldn't possibly be invented (unless the bureaucrat in Moscowwho controlled a given sector of the economy had a genuinely good idea, but that was rare).

Research, on the other hand, is much more prone to centralization, so it was doing quite well in the Soviets. Hence they ended up with a lot of research, but not that much innovation.

The article commented in this thread is one such example: neural networks were put to actual real-world use by western companies. Another example is the attempt of USSR to build something like the internet before the current one was thought out. They had the vision but the implementation somehow didn't take off.

Both the US and USSR knew about neural networks for decades before fast massively parallel hardware made them feasible long after the end of the USSR, so I'm not sure that's a useful comparison. The development of the internet was driven by the state out of national security concerns, not market forces and was not widely used during the USSRs life, so I'm not sure that one is helpful to illustrate the OPs point either.

I think this is incorrect, this isn't the first trained multi-layer neural network.

Minsky & Rosenblatt's https://en.wikipedia.org/wiki/Perceptrons_(book) covers multi-layer neural nets.

Rosenblatt's electronic implementation, the Mark I https://apps.dtic.mil/dtic/tr/fulltext/u2/236965.pdf

Minsky's electronic implementation, SNAC https://en.wikipedia.org/wiki/Stochastic_neural_analog_reinf...

I mentioned a “deep” multi-layer neural net, which implies more than just 2-3 layers. As far as I’m aware, this is the first eight-layer trained network mentioned in any publication. http://people.idsia.ch/~juergen/deep-learning-conspiracy.htm...

Ah, I wasn’t aware of that distinction.

Missing from the article is any discussion of control theory. In my view cybernetics essentially morphed into modern day control theory, there are many examples of very broad ideas in cybernetics being made specific and rigorous in controls. For example the good regulator theorem became the internal model principle. I encourage anyone interested in learning more about this brand of cybernetics to pick up a control or dynamical systems textbook, researchers have definitely not been idle in this area!

From article: "Over time, cybernetics’s contributions were stripped away and reclassified as something else. Early work on feedback loops in cognition became AI. Models of the format of feedback loops became information theory. Work on feedback loops around production became part of economics. Work on mathematical representations of complex systems became system dynamics. Work on feedback loops in behavior moved over to sociology and psychology. Cybernetic theory practically founded anthropology and cognitive science."

Control theory is just one amongst many splinters, without any implied primacy of control theory.

Control theory certainly predates Wiener's cybernetics. He did not invent negative feedback analysis nor design of control systems. He did fail to cite previous work, so adherents of cybernetics think he did.

In Cybernetics, he does credit Maxwell, Pitts, etc. He goes through the entire history of science relevant to Cybernetics for the first and second chapters.

He did invent plenty of feedback analysis techniques and modern control systems before writing Cybernetics

There are many developments relevant for cybernetics that were not mentioned in the book Cybernetics, like discoveries of H.S. Black and other engineers in Bell labs, or early mechanical analog computers. Here is a short review of book on the history of control engineering before cybernetics: https://onlinelibrary.wiley.com/doi/pdf/10.1002/acs.839

"[...] general reader could be forgiven for assuming that Wiener was the originator of the idea of feedback control. Significant though these war-time contributions were, they were not the whole story: as I and others have shown feedback control has had a much longer history"

That text includes more references to papers and books on the history of ideas relevant for control theory and control engineering.

Indeed. It is a necessarily true model of causality.

Highly recommend Minsky's application of the theory on computation: Computation: Finite and Infinite Machines

Interestingly, Minsky went out of his way to hide the production of that project without Weiner to avoid his influence on it--whom he disliked at least at a professional capacity.

There's some part where he shows the symmetry of the machine with the environment implies they are are interchangeable and is merely a description of convenience. Of course this trivially flows from the time reversibility.

here's the pdf: https://news.ycombinator.com/item?id=15621846

>The word cybernetics comes from the greek κυβερνήτης (kybernḗtēs) which means a pilot, governor, someone who steers.

TIL that's also where Kubernetes comes from.

Principia Cybernetica was one of the older sites on the web that I remember. Mirrored here: http://pespmc1.vub.ac.be/MST.html

That's a link to the 'metasystem transition', a concept coined by Russian cybernetician Valentin Turchin (not mentioned in the article).

Cybernetics sounds cool (always has), but I think it lives on nowadays as control theory. Feel free to correct me if I'm wrong. Seems like most control theory guys are working on autonomous vehicles now.

The concept of the metasystem transition has inspired me for a long time. In almost a spiritual sense. Because it just is such a powerful event. I imagine someday brain computer interfaces or general AI could give rise to a metasystem transition. Hopefully it won't turn out like those Star Trek episodes.

For some arcane reason, my CS degree had a course in Cybernetics and Systems Theory (which is weird because it is not taught in any other degree in my country, and it is taught by one professor that happened to study Cybernetics in the 60s in East Germany). It is possibly the one course that changed my perception about things, interactions and feedback loops, and it still piques my curiosity.

Probably because 'Cybernetics' has a way cooler ring to it than 'Control Systems'.

This may be why an 18-year old me chose to do a degree in Cybernetics and Control Engineering! Excited by the first word, I didn't dwell too much on the last two.

Sure, there was some fun robotics, electronics and math, but also some truly baffling control theory. Thankfully there was also a small amount of Computer Science, which sent me down the programming path that I remain on today...

Is there any point in learning about cybernetics these days? It certainly sounds very important. But does it have an impact on software engineering as we know it today?

Side note: I find it interesting that Weiner's book, which was fairly technical and full of rather arcane calculus, was widely popular among general readers. Just goes to show how much reading culture has changed since the 50s.

I would highly recommend reading Brain of the Firm by Stafford Beer.

It’s the most fantastic layman’s introduction to cybernetics, written from the perspective of the economist behind the “Cyber” in the Chilean CyberSyn project (1970-3).

I discovered him reading People’s Republic of Walmart, another fantastic book about how effective socialist strategies for centralized government are applied in piecemeal to the largest corporations on the globe.

Highly, highly recommended—-and yes, I believe it’s valuable to learn about cybernetics. It teaches us that we don’t need more computers, or faster computers—-we need to use computers in different ways. Basically, we’ve already got the solutions to our problems, but we don’t implement them because we’re trapped by classic civilizational barriers (it’s hard to change people’s minds in the face of a culture that doesn’t want their minds to change).

For a more recent study about the Chilean approach, I can recommend "Cybernetic Revolutionaries" by Eden Medina (2011, MIT Press, [link](https://mitpress.mit.edu/books/cybernetic-revolutionaries)).

+1 This is the standard work on Chile's Cybersyn. All articles and chapters on it found elsewhere are basically rewritings of this book.

Thank you for this!

Coincidentally, the market price of copper crashed in 1970 and stayed low until 1973, when it spiked. Compare https://news.ycombinator.com/item?id=24260354 or https://news.ycombinator.com/item?id=23908438 . Now I'm curious: does anyone have an example of a reform/revolution lucky enough to be selling exports into a rising market?

(Mid-1980s conservative economic reform in the US might be an example, although in that case I suspect the Plaza Accord had more to do with it than pure market forces: https://news.ycombinator.com/item?id=23219277 )

From what I understand, the economic revolution / experiment in Chile was abruptly ended due to political forces, not market forces.

Basically, any socialist (sympathizing) government during the under a US magnifying glass. Oil money went into funding far-right fascist / military groups, which eventually caused turmoil in a variety of industries, including a steel worker’s strike.

The entire CyberSyn operation was conducted on a series of glorified typerwriters—-at the time, they were the only computers that the Chilean government could purchase, due to sanctions.

If the US had stopped strangling the nascent Chilean economy, then it would have been better positioned to handle social turmoil and competing political interests.

Instead, on September 11th, 1973, far right militias backed by US money raided Allende’s central command—-he ended up committing suicide rather than ending up a political prisoner.

(Sorry if this was a bit unrelated to the initial reply—it just bugs me how removed from actual market forces these economies are operating in. I feel like it doesn’t matter what you are exporting—-if the US wanted it post-WW2, they came in and took it).

I haven’t read any of the books mentioned, but as a software worker I’m sure many of us have seen problems which would be solvable, even very simply, if the other parameters (often people) were more readily changeable.

I’ve done some reading on cybernetics but I’ll have to look up Beer.

Actually it was really The Human Use of Human Beings published two years after Cybernetics which was wildly popular, Wiener having rewritten his argument for a 'general audience' - sans all that calculus.

How well read was Wiener actually though? The historian Ronald Kline compares these books to the status of Steven Hawkins' Brief History of Time upon its release. The books anyone invested in intellectual development needed /to show/ they had a copy of, rather than one they exactly made it through cover to cover.

Does https://www.youtube.com/watch?v=K-yTrFDHpbw&t=330 demonstrate that the warnings of The Human Use of Human Beings have crossed the chasm from highbrow adult books to lowbrow children's cartoons over the last seven decades?

Minor point, but I had a fascinating lunchtime chat with a fellow who made his pile[1] back in the 1950's, because he saw I was reading Weiner's book.

[1] He developed automatic reset for bowling alleys, and made bank. For his second system, he thought he'd automate food service. That didn't go (and hasn't gone) so well.

Worth mentioning the very popular 1960 self-help book “Psycho-Cybernetics”, by Maxwell Maltz. Predates Doctor Who’s Cybermen as a mainstream use of the term.

https://en.wikipedia.org/wiki/Psycho-Cybernetics

cybernetics links:

Second-order cybernetics https://en.wikipedia.org/wiki/Second-order_cybernetics

Systems_theory: Cybernetics https://en.wikipedia.org/wiki/Systems_theory#Cybernetics

what is cybernetics? https://vimeo.com/41776276

design for a self-regenerating organization https://www.pangaro.com/ashby+design-for-self-regenerating-c...

notes 358 on the role of leadership & language in regenerating organizations http://www.pangaro.com/littlegreybook-dom.pdf

towards a theoretical foundation for agile development https://ieeexplore.ieee.org/document/7302798

2017 cybernetics conference https://livestream.com/internetsociety2/cybercon

2017 cybernetics conference - proxy politics as social cybernetics https://livestream.com/internetsociety2/cybercon/videos/1662...

how cybernetics connects computing, counterculture, and design http://www.dubberly.com/articles/cybernetics-and-countercult...

a tale of the mirror world, part 1: calculators and cybernetics https://www.filfre.net/2017/06/tales-of-the-mirror-world-par...

Niklas Luhmann: Systems Theory https://en.wikipedia.org/wiki/Niklas_Luhmann#Systems_theory

I first learned about cybernetics through the book Creation: Life and How to Make It by Steve Grand. Grand created the Creatures video game and applied ideas from cybernetics. As a result, each Norn (the name of the fictional species the game is about) has a genome, a neural network brain, a fairly complex biochemistry, sensors and actuators.

This article reminded me of Ferren Alet's work at MIT. I think it's been clear that feedback loops are important for certain types of intelligence for a long time. And its exciting to see things like lots of feedback loops and modularity being emphasized more in neural networks research.

Whoops. Its Ferran not Ferren.

Cybernetics is really an old word, or one of the old words, for AI/robotics.

The overarching, multidiscipline, that subsumes cybernetics, can be summarized here:

https://upload.wikimedia.org/wikipedia/commons/d/de/Complex_...

> Cybernetics is really an old word, or one of the old words, for AI/robotics.

Completely wrong.

AI/robotics are but two of the many fields that cybernetic principles apply to, and arguably not even the most important.

I suggest you look at what people like Bateson, Von Foerster and Glanville were doing, to understand first the scope and ultimately what cybernetics really is at its core.

Ah, this is so handy!

It explains why some Economics degrees have a course on Cybernetics.

> The word cybernetics comes from the greek κυβερνήτης (kybernḗtēs) which means a pilot, governor, someone who steers.

TIL: Where Kubernetes got its name from.

Reply on news.ycombinator.com