Chronaxie Infinity, vol 12: Why robot-eating robots are not cannibals
A case study in the insidiousness of the doomer narrative. Plus, mollicutes!
Happy Tuesday, everyone. Let’s talk about robot cannibals.
I’m unacceptably late to this spate of July news stories about robot cannibalism, but if you’re having the same experience of time and the world that I’m having, July may as well be ancient Greece. Even if you saw this study when it was published, the density and stupidity of subsequent events may have wiped it from your mind.
Couple reasons I bring this up now. First, if you did miss it, the problem of how to create self-repairing, self-growing machines is super interesting in itself. In July, the journal Science published a new approach to solving it — a framework that would enable a machine to integrate parts of other machines to fix or grow itself. But the reason I still want to talk about it now is because of the way it was headlined. Hod Lipson, one of the engineers who authored the Science paper, had all kinds of interesting things to say about this bioinspired approach to self-repair, including the alluringly alliterative, hot-to-go phrase “machine metabolism”. But no one put that in their headline. Instead, every outlet went full cannibal. (The clear winner was “The Cannibal Machines Are Coming—and They’re Evolving Without You”.)
I thought this was an interesting example of how a neutral development gets sucked into a pre-existing narrative. (And yes, yes, many of those headlines were tongue-in-cheek, and I understand, as a journalist, that headlines are supposed to get attention. Also the stories themselves were good.)
But why did so many editors simultaneously grok that Robot Cannibals was such a lock-in for attention?
It’s because when they work, good headlines surf a sea of vibes and shared assumptions. And right now, one of our few remaining shared assumptions is that we live in a dystopian hellscape, in which of course some crazy asshole is building cannibal robots. Only after clicking through or reading the subheadline — and by the way very few people even get that far, despite the most desperate efforts of the editors — would you understand that this is a story about a clever engineering approach. The vast majority of eyeballs simply coasted past “Robot Cannibals”, grimly nodded and thought “yep, tracks”, and moved to the next item on the doomscroll.
In fact, the paper has nothing to do with cannibalism — unless you think all animals are cannibals. We all radically disassemble other living things and force them to become part of us. Eating plants and other animals fixes the carbon that creates and maintains our physical form. Machine metabolism is about getting machines to do what living things have already been able to do for millions and millions of years.
There’s a better word for this than cannibalism, and by “better” I mean “Headline Kryptonite”. It comes out of the 1970s, when the biologist-philosophers Humberto Maturana and Francisco Varela coined an elegant little term that uniquely describes the quality that makes life so special: autopoiesis.
Autopoiesis is the essence of why it is so hard to get machines to do what life does: life has a special autonomy that just can’t be copied in circuits. One famous example that drove it home for me is the thought experiment on what an airplane would have to be like if it were alive.
If you wanted to build an airplane that worked like a living organism, it would have to be able to repair or regrow its own parts in mid-flight if something broke; it would react to environmental surprises, such as storms, by spontaneously making choices; and it would hunt and ingest its own fuel.
Animals do all of this just as a consequence of existing, whereas a machine needs the most micromanagerial hand-holding: it is programmed, prompted, pandered to and provided for at every step. This is what’s actually intriguing about what Lipson and his colleagues at Columbia Engineering and the University of Washington did. They created the first step of a framework that could lead to machine autopoiesis.
Other scientists explore different ways to get there. At Tufts, Michael Levin is approaching the idea from exactly the opposite direction — instead of finding ways to make machines act like life, he is turning living matter into machine parts. Xenobots and anthrobots, respectively balls of frog- and human-derived cells (themselves already endowed with autopoietic properties), can be separated from their larger forms, and then coaxed into new assemblages that do work.
One question raised by all this work: where is the line between giving machines more autonomy so you don’t have to micromanage all their functions, and keeping them from acting on their own instincts if those don’t align with your priorities? Well. Maybe we’re back to robot cannibals after all.
Some quick links
“In winter 2021, a massive power outage hit Texas. Investigations ultimately found that the outage’s cause was gas and coal plants that were forced offline because of freezing temperatures. But critics of renewable energy blamed solar and wind, says Timmons Roberts, professor of environmental studies and sociology at Brown University.” (NPR)
Speaking of bad narratives, this excellent article at NPR takes on the pernicious tendency to blame big blackouts on renewables. The latest episode concerns the fiasco that took down the Iberian grid in April. Which, as a new report makes plain, was not caused by renewables at all — so reporter Julia Simon dug into why petrol interests love to spread that lie after every blackout.“These days, we are all workers in the internet factory: We supply the images, write the copy, engage with the ads, promote the products. That labor sustains an inequitable tech economy in which a handful of companies, including Meta, Google and Amazon, rake in profit and enrich their shareholders. In return, consumers get a product designed to keep us scrolling.” (CNN)
Good CNN story on the growing new Luddite movement, and how normies like us are getting tired of being constantly exposed to surveillance and artificially ginned-up emotional friction. One day we just might up and boycott the whole thing. No problem, says Mark Zuckerberg (whose face appears in the dictionary under the phrase “This Fucking Guy”). He’ll just fill the empty space with AI slop.“One friend told me about her insomniac mother who regularly reels off “rot” she’s read on Facebook at 4am, and the endless useless packages she can’t afford that arrive from Instagram ads. Another described the ongoing struggle she has with the way her mother-in-law is fixed on her phone when babysitting her grandson, and her concerns at the endless Trump speeches served by her elderly dad’s YouTube algorithm.”
We’re marinating in messages about the dangers of phones and social media and tech for our vulnerable youth. Not a lot of ink has been spilled about “the over-70s who are second only to those in their 20s when it comes to the average amount of time spent online.” Everyone I know with older parents has been increasingly, uneasily aware of this. This nice writeup on Eva Wiseman’s substack, The Pyjama Monologues, lays out the problem.From the Department of Ridiculous Words in Science: Do Mollicutes Bleb?
Meet the mollicutes. This group of bacteria, whose name means soft skin, derive their name from their lack of a cell wall. (Where the vast majority of bacteria possess these protective tough walls, mollicutes are only encased in a flexible membrane, similar to your cells and mine.)
Probably because of this vulnerability, mollicutes don’t live in the wild. Instead they hitchhike inside the cells of animals and plants, and while they cause lots of economic and health chaos, the infections they cause there are rarely fatal.
Like all cells, they engage in a behaviour called “membrane blebbing” (I thought it was a typo when I first read it. It’s not. There is a paper out there whose title is “Blebs and former blebs”). Blebs — also known as snouts — are when bulges form in the cell membrane.
These blebs sometimes pop — that is, they cleave off from the originating cell and go float around the extracellular space. All existing cells release these so-called “extracellular vesicles”. They’ve been compared to delivery trucks, and sometimes the cargo they carry is good (signalling molecules) and sometimes it is bad (cancer stuff). I like to think of them as the speech bubbles that make cell-to-cell communication possible.
Anyway a longstanding problem for mollicute blebs has been that they are so tiny that it’s been hard to distinguish their blebs from themselves. Now researchers have maanged to characterise the extracellular vesicle cargo.
But what’s really important is that the word “bleb” is now trapped in your head for eternity. It has been blebbed into your head by way of this Substack. You’ll be brushing your teeth or making coffee or looking at a tree and there it will be, unannounced and unwanted: “bleb”. It could be worse. It could be autopoiesis.
Blebs,
Sally
Art sources:
I use Public Work, which surfaces artwork in the public domain. The artwork illustrating robot cannibals above is from Mary Mills Lyall, and Earl Harvey Lyall, The Cubies’ ABC (New York: G. P. Putnam’s Sons, 1913)
The mollicute: Rainbow-coloured Beasts from 15th-Century Book of Hours