Monochrome
and the connections between meaningmaking, innovation, and not-knowing.
Hello folks,
The world creeps toward Spring, and direct sunlight now makes a brief but much-appreciated appearance every day in the office. This issue is about what it means to be new. Though I know it sounds weird, monochromes that have deeper meaning than they appear to are what connects a bunch of recent and not-so recent work I’ve been doing around newness, data, and meaning-making.
But first!
The mind-body nexus
The next discussion in my InterIntellect series about not-knowing is about the mind-body nexus and why it makes not-knowing feel so hard (“feel” in the sense of affect and emotion). That happens next Thursday, 16 March, 8-10p CET / 12-2p Pacific / 3-5p Eastern / 7-9pm UK, and is open to all (if the $15 ticket price is undoable, drop me a line and I’ll sort you out). You can get session info and tickets here. The short text to read in advance is called “Why not-knowing feels so hard.”
Two points about not-knowing
Earlier this week, I went up to Paris for the day to have dinner with an old pal passing through France. Of course, the “real” reason was to give a talk at SciencesPo. The talk was about untangling the connections between not-knowing and innovation given the particular context of public sector work. The argument I made was that:
Innovation requires a productive relationship with not-knowing.
Productive relationships with not-knowing require clarity about what not-knowing is.
Public sector wicked problems inevitably require innovative solutions.
Public sector organisations are designed to impede productive relationships with not-knowing, which impedes innovation.
Sneaky strategies can circumvent these design impedances and enable public sector innovation.
Two points about not-knowing came up in the discussion. They’re relevant even outside public sector work and also explain my current focus on figuring out how to think clearly about different types of not-knowing.
A big fear of acknowledging not-knowing explicitly is that this will result in losing legitimacy and authority. This is a big deal in organisations (whether private or public). The way to get around this is to invest effort in explaining what you know, what you don’t know, and what plans you have to figure things out. This is a discipline which reflects better thinking and produces better action, and shows the ability to navigate situations of uncertainty with honesty. A simpler way to say it is: Being able to acknowledge not-knowing is not the same thing as not knowing anything.
There’s an understandable bias for action and “not wasting time on theory” (unless you’re an academic, then the bias is for “rigorous theory” even if utterly disconnected from practice). The separation between the two is a misconception. We can only choose actions properly if we are thinking clearly about the world in which we take those actions. This is why there are serious consequences to fuzzy thinking about not-knowing. We simply cannot act better without thinking more clearly, or think more clearly without learning from action. A simpler way to say it is: In relation to not-knowing, thinking clearly and acting better are two sides of the same coin.
I’m working on writing up the talk and cleaning up the audio recording; will post both in the next few weeks.
Meaning-making, not-knowing, innovation, data
Last week, some neat connections emerged between several of my seemingly disparate threads of enquiry — on meaning-making as a human capacity, on not-knowing, on innovation and newness, and on why data should not be allowed to speak for itself.
It started because I went to the opening show of a new ephemeral curatorial project here in Marseille called Xanadu.
Christophe Bruno and Jeff Guess (two artists in this year’s cohort of IMERA Fellows) inaugurated Xanadu last Thursday with a speculative work by Gwenola Wagon and Pierre Cassou-Noguès exploring how technology might condition our imagination of housing and the search for habitation — by using LLMs to generate both real estate listings and photos.
At the opening, Christophe, Jeff, and I accidentally discovered that we’re all currently exploring aspects of not-knowing. We met again on Friday to talk more. Some time that afternoon, a paper by Chris Calude and Giuseppe Longo on the inevitability of spurious correlation in big datasets came up in conversation.
Whenever someone builds a big data set, there’s some assertion that somehow the bigness of the data means that answers that come out of the dataset are really good and that theorising — i.e. meaning-making in relation to the interrogation and evaluation of data — is therefore no longer needed. It’s a thinly disguised version of the apparently unkillable belief that “the data speaks for itself,” and one of the reasons for ever-larger datasets in a wide range of research.
The latest incarnation of this kind of thinking shows up in how we’re talking about the increasingly enormous corpuses of data on which LLMs (large language models) are trained. There seems to be a growing belief that these LLMs can now do what humans can do. I disagree, of course.
To me, these LLMs can at best imitate meaning that humans have previously created. Things can generate information (e.g., a sensor can generate a stream of temperature readings for a region), but only humans can generate meaningful information (e.g. the stream of temperature readings for a region means that dangerous and undesirable anthropogenic climate change is happening).
Because art practice is not primarily functional, it is one of the settings in which humans most clearly create meaningful information. The connection to not-knowing is that it is a necessary precondition for producing meaningful new information.
And so we get to monochromes. The example Christophe brought up was of monochromes in art, specifically the monochromes of Alphonse Allais1 which pre-dated the better known monochrome works of Malevich and others. He’s been trying, without success, to prompt an LLM to generate a monochrome in response to Allais’s prompts.2
One of Allais’s monochromes is below. The title or prompt, loosely translated, is “Apoplectic cardinals harvesting tomatoes by the Red Sea.” The creation of new meaningful information here is in Allais defining something (a red rectangle) as a representation of something which it does not superficially appear to be (red-faced cardinals, tomatoes, the Red Sea). Creating this new conceptual connection is an act of meaning-making.
The first artist to make a thing more than what it appears to be has to do this from a position of not-knowing whether this action is possible. In other words, they must be in a situation where they don’t know if this action of applying a specific but new meaning will be read as an artistic act. Artists who come after (like Malevich) have the advantage of knowing that this action is possible and their task of creating new artistic actions then requires putting themselves into other situations of not-knowing.
Novelty only comes from entering into situations of not-knowing and resolving them (whether permanently or temporarily) by creating new meaning. This is not easy and the machines can’t do it (yet).
So…
Writing, recent and not-so-recent
I wrote recently on why meaning-making is what sets us apart from the machines (for now).
And several years ago on why data shouldn’t be allowed to speak for itself.
See you soon,
Vaughn
Next time in Honfleur, I will for sure visit the Alphonse Allais Museum (“De plus, souvenons-nous qu’au livre des records, ce musée est le plus petit musée du monde. Cela est un plus dans sa valorisation”) — it sounds spiritually aligned with the Museum of Jurassic Technology, not far from where I lived in LA.
There may also be other reasons for this, to do with the content of the training corpus of the LLM he has been using.