Beyond the Age of Machines
This week, I was invited to give a three-minute flash talk at an event called "Human Development, Sustainability, and Agency," which was organized by IIASA (the International Institute for Applied Systems Analysis), the United Nations Development Programme (UNDP), and the Austrian Academy of Sciences (ÖAW). The event framed the release of an UNDP report called "Unsettled times, unsettled lives: shaping our future in a transforming world." It forms part of IIASA's "Transformations within Reach (TwR) project, which looks for ways to transform societal decision-making systems and processes to facilitate transformation to sustainability.
You can find more information on our research project on agency and evolution here.
My flash talk was called "Beyond the Age of Machines." Because it was so short, I can share my full-length notes with you. Here we go:
"Hello everyone, and thank you for the opportunity to share a few of my ideas with you, which I hope illuminate the topic of agency, sustainability, and human development, and provide some inspiring food for thought. I am an evolutionary systems biologist and philosopher of science who studies organismic agency and its role in evolution, with a particular focus on evolutionary innovation and open-ended evolutionary dynamics. I consider human agency and consciousness to be highly evolved expressions of a much broader basic ability of all living organisms to act on their own behalf. This kind of natural agency is rooted in the peculiar self-manufacturing organization of organisms, and the consequences this organization has on how organisms interact with their environment (their agent-arena relationship). In particular, organisms distinguish themselves from non-living machines in that they can set and pursue their own intrinsic goals. This, in turn, enables living beings to realize what is relevant to them (and what is not) in the context of their specific experienced environment. Solving the problem of relevance is something a bacterium (or any other organism) can do, but even our most sophisticated algorithms never will. This is why there will never be any artificial general intelligence (AGI) based on algorithmic computing. If AGI will ever be generated, it will come out of a biology lab (and will not be aligned with human interests), because general intelligence requires the ability to realize relevance. And yet, we humans increasingly cede our agency and creativity to mindless algorithms that completely lack these properties. Artificial intelligence (AI) is a gross misnomer. It should be called algorithmic mimicry, the computational art of imitation. AI always gets its goals provided by an external agent (the programmer). It is instructed to absorb patterns from past human activities and to recombine them in sometimes novel and surprising ways. The problem is that an increasing amount of digital data will be AI-generated in the near future (and it will become increasingly difficult to tell computer- and human-generated content apart), meaning that AI algorithms will be trained increasingly on their own output. This creates a vicious inward spiral which will soon be a substantial impediment to the continued evolution of human agency and creativity. It will be crucial to take early action towards counteracting this pernicious trend by proper regulations, and a change in the design of the interfaces that guide the interaction of human agents with non-agential algorithms. In summary, we need to relearn to treat our machines for what they are: tools to boost our own agency, not masters to which we delegate our creativity and ability to act. For continued sustainable human development, we must go beyond the age of machines. Thank you very much."
SOURCES and FURTHER READING:
"organisms act on their own behalf": Stuart Kauffman, Investigations, OUP 2000.
"the self-manufacturing organization of the organism": see, for example, Robert Rosen, Life Itself, Columbia Univ Press, 1991; Alvaro Morena & Matteo Mossio, Biological Autonomy, Springer, 2015; Jan-Hendrik Hofmeyr, A biochemically-realisable relational model of the self-manufacturing cell, Biosystems 207: 104463, 2021.
"organismic agents and their environment": Denis Walsh, Organisms, Agency, and Evolution. CUP, 2015.
"the agent-arena relationship": a concept first introduced in John Vervaeke's "Awakening from the Meaning Crisis," and also discussed in this interesting dialogue.
"agency and evolutionary evolution": https://osf.io/2g7fh.
"agency and open-ended evolutionary dynamics": https://osf.io/yfmt3.
"organisms can set their own intrinsic goals": Daniel Nicholson, Organisms ≠ Machines. Stud Hist Phil Sci C 44: 669–78.
"to realize what is relevant": John Vervaeke, Timothy Lillicrap & Blake Richards, Relevance Realization and the Emerging Framework in Cognitive Science. J Log Comput 22: 79–99.
"solving the problem of relevance": see Standford Encyclopedia of Philosophy, The Frame Problem.
"there will never be artificial general intelligence based on algorithmic computing": https://osf.io/yfmt3.
"we humans cede our agency": see The Social Dilemma.
So, this is as good a reason as any to wake up from my blogging hibernation/estivation that lasted almost a year, and start posting content on my web site again. What killed me this last year, was a curious lack of time (for someone who doesn't actually have any job), and a gross surplus of perfectionism. Some blog posts got begun, but never finished. And so on. And so forth.
So here we are: I'm writing a very short post today, since the link I'll post will speak for itself, literally.
I've had the pleasure, a couple of weeks ago, to talk to Paul Middlebrooks (@pgmid) who runs the fantastic "Brain Inspired" podcast. Paul is a truly amazing interviewer. He found me on YouTube, through my "Beyond Networks" lecture series. During our discussion, we covered an astonishingly wide range of topics, from the limits of dynamical systems modeling, to process thinking, to agency in evolution, to open-ended evolutionary innovation, to AI and agency, life, intelligence, deep learning, autonomy, perspectivism, the limitations of mechanistic explanation (even the dynamic kind), and the problem with synthesis (and the extended evolutionary synthesis, in particular) in evolutionary biology.
The episode is now online. Check it out by clicking on the image below. Paul also has a break-down of topics on his website, with precise times, so you can home in on your favorite without having to listen to all the rest.
Before I go, let me say this: please support Paul and his work via Patreon. He has an excellent roster of guests (not counting myself), talking about a lot of really fascinating topics.
Hello everybody. This is my first blog post. I was undecided at first. What do I write about? Where do I begin? Then, last night, I came across this article by Michael Levin and Daniel Dennett in Aeon Magazine. It illustrates quite some of the problems—both in science and about science—that I hope to cover in this blog.
"Cognition all the way down?" That doesn't sound good... and, believe me, it isn't. But where to begin? This article is a difficult beast to tackle. It has no head or tail. Ironically it also seems to lack purpose. What is it trying to tell us? That cells "think"? Maybe even molecules? How is it trying to make this argument? And what is it trying to achieve with it? Interdisciplinary dialogue? Popular science? A new biology? I think not. It does not explain anything, and is not written in a way that the general public would understand. I do have a suspicion what the article is really about. We'll come back to that at the end.
But before I start ripping into it, I should say that there are many things I actually like about the article. I got excited when I first saw the subtitle ("unthinking agents!"). I'm thinking and writing about agency and evolution myself at the moment, and believe that it's a very important and neglected topic. I also like the authors' concept of teleophobia, an irrational fear of all kinds of teleological explanations that circulates widely, not only among biologists. I like their argument against an oversimplified black-and-white dualism that ascribes true cognition to humans only. I like their call for biologists to look beyond the molecular level. I like that they highlight the fact that cells are not just passive building blocks, but autonomous participants busy building bodies. I like all that. It's very much in the spirit of my own research and thinking.
But then, everything derails. Spectacularly. Where should I start?
AGENCY ISN'T JUST FEEDBACK
The authors love to throw around difficult concepts without defining or explaining them. "Agency" is the central one, of course. From what I understand, they believe that agency is simply information processing with cybernetic feedback. But that won't do! A self-regulating homeostat may keep your house warm, but does not qualify as an autonomous agent. Neither does a heat-seeking missile. As Stuart Kauffman points out in his Investigations, autonomous systems "act on their own behalf." At the very least, agents generate causal effects that are not entirely determined by their surroundings. The homeostat or missile simply reacts to its environment according to externally imposed rules, while the agent generates rules from within. Importantly, it does not require consciousness (or even a nervous system) to do this.
AGENCY IS NATURAL, BUT NOT MECHANISTIC
How agents generate their own rules is a complicated matter. I will discuss this in a lot more detail in future posts. But one thing is quite robustly established by now: agency requires a peculiar kind of organisation that characterises living systems—they exhibit what is called organisational closure. Alvaro Moreno and Matteo Mossio have written an excellent book about it. What's most important is that in an organism, each core component is both producer and product of some other component in the system. Roughly, that's what organisational closure means. The details don't matter here. What does matter is that we're not sure you can capture such systems with purely mechanistic explanations. And that's crucial: organisms aren't machines. They are not computers. Not even like computers. Rosen's conjecture establishes just that. More on that later too. For now, you must believe me that "mechanistic" explanations of organisms based on information-processing metaphors are not sufficient to account for organismic agency. Which brings us to the next problem.
EVOLVED COMPUTER METAPHORS
We've covered quite some ground so far, but haven't even arrived at the main two flaws of the article. The first of these is the central idea that organisms are some kind of evolved information-processing machines. They "exploit physical regularities to perform tasks" by having "long-range guided abilities," which evolved by natural selection. Quite fittingly, the authors call this advanced molecular magic "karma." Karma is a bitch. It kills you if you don't cooperate. And here we go: in one fell swoop, we have a theory of how multicellularity evolved. It's just a shifting of boundaries between agents (the ones that were never explained, mind you). Confused yet? This part of the article is so full of logical leaps and grandstanding vagueness that it's really hard to parse. To me, it makes no sense at all. But that does not matter. Because the only point it drives at is to resuscitate a theory that Dennett worked on throughout the 1970s and 80s, and which he summarised in his 1987 book The Intentional Stance.
THE INTENTIONAL STANCE
The intentional stance is when you assume that some thing has agency, purpose, intents in order to explain it, although deep down you know it does not have these properties. It used to be big (and very important) in the time when cognitive science emerged from behaviourist psychology, but nowadays it mostly applies to rational choice theory applied in evolutionary biology. For critical treatments of this topic, please read Peter Godfrey-Smith's Darwinian Populations and Natural Selection, and Samir Okasha's Agents and Goals in Evolution. Bottom line: this is not a new topic at all, and it's very controversial. Does it make sense to invoke intentions to explain adaptive evolutionary strategies? Let's not get into that discussion here. Instead, I want to point out that the intentional stance does not take agency serious at all! It is very ambiguous about whether it considers agency a real phenomenon, or whether it uses intentional explanations as purely heuristic strategy that explicitly relies on anthropomorphisms. Thus, after telling us that parts of organisms are agents (at least that's how I would interpret the utterly bizarre "thought experiment" about the self-assembling car) they kind of tell us now that it's all just a metaphor, this agency thing. What is it, then? This is just confusing motte-and-bailey tactics, in my opinion.
AGENCY IS NOT COGNITION!!!
So now that we're all confused whether agency is real or not, we already get the next intellectual card trick: agency is swapped for cognition. Just like that. That's why it's "cognition all the way down." You know, agency is nothing but information processing. Cognition is nothing but information processing. Clearly they must be the same. There's just a difference in scale in different organisms. Unfortunately, this renders either the concept of agency or the concept of cognition irrelevant. Luckily, there is an excellent paper by Fermín Fulda that explains the difference (and also tells you why "bacterial cognition" is really not a thing). Cognition happens in nervous systems. It involves proper intentions, the kind you can even be conscious of. Agency, in the broad sense I use it here, does not require intentionality or consciousness. It simply means that the organism can select from a repertoire of alternative behaviours when faced with opportunities or obstacles in its perceived environment. As Kauffman says, even a bacterium can "act on its own behalf." It need not think at all.
PANPSYCHISM: NO THANK YOU
By claiming that cells (or even parts of cells) are cognitive agents, Levin and Dennett open the door for the panpsychist bunch to jump on their "argument" as evidence for their own dubious metaphysics. I don't get it. Dennett is not usually sympathetic to the views of these people. Neither am I. Like ontological vitalism, panpsychism explains nothing. It does not explain consciousness or how it evolved. Instead, it explains it away, negating the whole mystery of its origins by declaring the question solved. That's not proper science. That's not proper philosophy. That's bullshit.
SO: WHAT'S THE PURPOSE?
What we're left with is a mess. I have no idea what the point of this article is. An argument for panpsychism? An argument for the intentional stance? Certainly not an argument to take agency serious. The authors seem to have no interest in engaging with the topic in any depth. Instead, they take the opportunity to buzzword-boost some of their old and new ideas. A little PR certainly can't harm. Knowing Michael Levin a little by now, I think that's what this article is about. Shameless self-promotion. Science in the age of selfies. A little signal, like that of the Trafalmadorians in The Sirens of Titan that constantly broadcasts "I'm here, I'm here, I'm here." And that's bullshit too.
To end on a positive note: the article touches on a lot of interesting topics. Agency. Organisms. Evolution. Philosophical biology. Reductionism. And the politics of academic prestige. I'll have more to say about all of these. So thank you, Mike and Dan, for the inspiration, and for setting such a clear example of how I do not want to communicate my own writing and thinking to the world.
Life beyond dogma!