Commonplace book: far too much AI but maybe also some hope

First, calendars. This is of interest as I find my way around the communities in my new area. 

LAUTI has written a generous how-to on running a community calendar — the hard work of keeping a list of what's happening locally. It's all about the people, and talking to people is step 1. You don't have to roll your own software as they did; Mobilizon is a federated solution, or there's Gancio ('a shared agenda for local communities', AGPL). Martin Hamilton noted the gap for 'weird nerd' type event listings and the thread up and down includes some other maker-world examples. 

A new local friend for me is the Liverboard.  Lots of lovely things - almost too many. Social Liverpool's email events list has been good too.  DoES Liverpool also has an events system (as well as the maker calendar), which is good except you don't get an email confirmation, which has somehow become part of the events culture that my brain expects, leaving me baffled.  

In contrast to the usual torrent of bad news around tech:

Google didn't let you block AI overviews etc without also blocking search. Now that's changing: Google will let publishers opt out of AI summarisation features affecting search, driven by pressure from the CMA in the UK

I had not spotted this until it came up in Ben Evans's newsletter. Well done CMA! 

Tom Scott writes about the migration of the new — on how the internet is wonderful at helping people find each other, but weaker at catalysing something new. 

The internet satisfies the need before it becomes a movement. [...] Belonging and rupture are different things. Much of contemporary digital culture feels less like a movement than a club.

A movement turns not-belonging into a world. A platform turns not-belonging into a recommendation.

The new has not disappeared. It has been domesticated.

And on feeds as an example: 

... A feed has no real edge. It turns culture into a vertical surface, ordered by a system you cannot see and refreshed by a gesture so small it feels like instinct. It is the same gesture whether you are looking at friends, politics, work, shopping, films, music or jokes. Scroll, pause, continue. This feels neutral, but it teaches a theory of culture. Whatever matters will come to you, in sequence, if you keep moving.

There are benefits. Interfaces are easier to learn. Services are more accessible. The web is less hostile to ordinary users than it once was. There are good reasons for consistency.

But consistency is not culturally neutral. It trains expectation. It rewards the already-known. It makes deviation feel like failure. A strange interface is now often understood as a bad interface, rather than as a possible world.

 So we are losing things, in this platform world:

Fun is not the same as engagement. Play is not the same as retention. Beauty is not the same as polish. Surprise is not the same as novelty content. Meaning is not the same as relevance.

Engagement wants the next click, the next episode, the next level, the next session. Play may want the opposite. It may want pause, waste, difficulty, silence, wandering, boredom, failure, repetition without reward, or a joke that only works once.

... And perhaps the real loss is not originality, but pressure. The pressure that comes from not finding your people. The pressure that comes from having no ready-made category. The pressure that comes from needing to invent a form before you can be recognised.

A community lets you belong to the world as it is.

A movement asks you to make another one.

Good perspective - thanks Mia Ridge for sharing the link.

René Walter's "skeleton library" (via Sentiers) is a way of thinking about LLMs through the history of archives:

In the first of four parts on what he calls “interpolatable archives,” René Walter builds a conceptual framework for understanding LLMs through the history of archival thought. He opens with Aby Warburg’s Verknüpfungszwang—a compulsion to connect—and traces a line through Borges’s infinite library and Kenneth Goldsmith’s typewritten copy of Western literature to the training corpora of contemporary AI. The common thread is the human drive to find meaning inside vast, unruly accumulations of knowledge, and the successive architectures people have built to make that possible.

The piece centers on Walter’s “skeleton library” metaphor. An LLM, he argues, is not a library, it is what remains after a library’s contents have been atomised into a geometry of vectors. The texts are gone; what survives is a structure of statistical relationships so granular that any content can be reconstructed, blended, or redirected by warping the architecture itself. Prompting, in his reading, is not retrieval but morphing: bending the cookbook shelf into crime fiction, folding punk lyrics into classical literature. The output is not a document from the collection but an interpolation across it, a “chimera” that never existed in the original archive.

He closes the essay by framing LLMs as a new form of orality rather than literacy. Where traditional archives ground knowledge in traceable sources and authorial intent, these interpolatable archives generate responses the way pre-Homeric bards constructed epic verse: on demand, from mnemonic formulas, without a fixed text behind them. Walter doesn’t dismiss the risks—knowledge untethered from its sources, answers generated without traceable grounding in fact—but he holds open the possibility that Warburg’s associative, non-linear archive was doing something conceptually similar a century ago, and that there is something potentially generative in that kind of navigation, not just something lost.

Less optimistically, the Guardian on what talking to machines does to how we talk:

In the same way autocomplete has increased how much we use the 1,000 most common words in our vocabulary, talking with chatbots and reading AI-generated text may further constrict our speech. A recent University of Coruña study found that machine-generated language has a narrower range of sentence length, averaging 12–20 words, and a narrower vocabulary than human speech. Machine-generated text reads as smooth and polished, but it loses the meanders, interruptions and leaps of logic that communicate emotion.

A 2022 study found that children in households that used voice commands with tools like Siri and Alexa became curt when speaking with humans, often calling out "Hey, do X" and expecting obedience, especially from anyone whose voice resembled the default-female electronic voices.

I was glad to spot "AI centrism" - a new coinage and very useful. Martha Lincoln has a piece on it:

A position that claims to split the difference between genAI boosterism and detraction, but still leads to its acceptance and use.

The worked example in this Bluesky thread is the CSU spending $55 million over four years on a commercial chatbot whose educational benefits are unproven: "As the administration has claimed, “AI is here to stay” and is “fundamentally changing how people learn, work and solve problems.” As their logic goes, the CSU thus has no choice: "W/o any specifics, the CSU admin unfailingly claims they’ve adopted AI “responsibly,” “thoughtfully,” and “ethically”"  

Via John Naughton, a surprisingly thoughtful and grounded speech from Matthew Sanders, speaking to the Catholic bishops on AI and the crisis of meaning (earlier, and much shorter, than the Encyclical on AI):

Your parishioners aren't in danger of believing the machine is God. They're in danger of forgetting that they aren't machines.

He includes some powerful examples of what this might mean for pastoral care, the sorts of issues parishioners will start to bring, eg "Teenagers in confession describing relationships with AI companions."

For two hundred years, the modern world has answered the question “Who are you?” with a reductive “What do you do?” The Industrial Revolution tied human dignity, quietly but ruthlessly, to economic output. We’ve lived inside what I call the GDP Era. And we’re now, in real time, watching that era end.

Automation is coming for knowledge work through agentic AI. Automation is coming for physical work through embodied AI. There’s no sanctuary. For the first time in human history, generating massive economic value won’t require massive amounts of human labour.

... If the Church’s response is to argue that human beings are still economically necessary, we’ll lose the argument. The response has to be more radical. The response has to be a refusal of the premise...that a person’s worth was ever economic in the first place.

There’s a political edge to this, and I think it needs to be named in this room, because no one else is going to name it. Historically, the working class’s ultimate leverage against the elite was the strike — the threat to withdraw labour. When labour is no longer necessary for production, that leverage vanishes. If the intelligent machines are owned by a small number of corporations, and the masses depend on a universal basic income paid for out of the taxes on those corporations, we haven’t built a liberation. We’ve built a digital feudalism — a society of dependents, not citizens. A Universal Basic Income in that configuration isn’t freedom. It’s an allowance.

He ends with a call for the church to build its own AI infrastructure(!).

Good to see the examples of alternative AI systems in Nathan's great article for the Conversation:

...the current pope has repeatedly invoked the earlier Leo’s 1891 encyclical Rerum Novarum. That document, which waded into the political and economic debates of the time, denounced the excesses of the Gilded Age and pointed toward a more just social order. 

... Leo XIII rejected both unfettered capitalism and revolutionary socialism. He invoked the medieval guilds, in which craftspeople self-organized, and asserted the rights of industrial workers to organize as well.

... The spirit of Rerum Novarum continued to spread. Starting in the 1950s, the largest network of worker cooperatives in the world, the Mondragon Corporation in Spain’s Basque region, was founded by a Catholic priest. It was a direct result of Leo XIII’s encyclical.

I had no idea!  

...  the encyclical calls on people everywhere to adopt “the pressing duty to remain profoundly human” – to be neither “spectators” nor “commentators” but to take an active role by participating in what he calls “the construction sites of history.” Some already are.

... In Switzerland, a collaboration between government and academia has produced Apertus, a foundational model based on fully documented designs and data sources – a far cry from the opaque and at times illegal practices of leading AI companies. Some of Apertus’ developers have created a consumer cooperative, enabling users to co-own their interface with the model.

Cooperative ownership like this allows users to tune AI experiences more intentionally toward their needs. The large U.S. farmer co-op Land O’Lakes, for example, has created AI-enabled tools that provide analysis and guidance for its members based on the data that they collectively co-own. The more nascent Transkribus in Europe is co-owned by research institutions that collectively train their AI software to transcribe texts for historical research. These kinds of systems follow Leo XIV’s call to “manage data as a common or shared good.”

Then there's destroysaas — a cooperative where small businesses collectively fund, own and govern the software they depend on; "not open source, not closed source, our source." The shifting software cost dynamics might have unexpected side effects.  Interested to see how that goes. 

Are we sick of AI yet? Sorry sorry. 

Artificial intelligence is becoming a bit like alcohol.

The cause of, and solution to, all of life’s problems.

Not my line. Homer Simpson’s. But it may be the most accurate description of this entire AI moment.

Thanks Martha for this clarity... 

... the next challenge in AI is not simply capability. It is legitimacy.

[examples]... I do not read this as anti-technology sentiment.

I think it reflects something deeper: a growing suspicion that the gains from this next technological wave will once again flow disproportionately to a relatively small number of people, firms and places.

 ... It is valuable to understand the history because AI is arriving into societies where trust in institutions is already fragile. ... Societies can absorb astonishing amounts of technological disruption when people believe the rules are legitimate and the gains are shared. They become far less tolerant when change feels extractive, opaque or designed somewhere else by people they neither know nor trust.

Bruce Schneier asks "How Dangerous Is Anthropic's Mythos AI?" - some clear thoughts on the obvious software security stuff, but mostly interesting for other things:

Just as these models are finding hundreds of vulnerabilities in complex software systems, we should expect them to be equally effective at finding many new and undiscovered tax loopholes. I am confident that the major investment banks are working on this right now, in secret. They’ve fed AI the tax code of the US, or the UK, or maybe every industrialized country, and tasked the system with looking for money-saving strategies. How many tax loopholes will those AIs find? ....

Sure, the AIs will come up with a bunch of tricks that won’t work, but that’s where those attorneys and accountants come in—to verify, and then justify, the loopholes. And then to market them to their wealthy clients.

As goes the tax code, so goes any other complex system of rules and strategies. These models could be tasked with finding loopholes in environmental rules, or food and safety rules—anywhere there are complex regulatory systems and powerful people who want to evade those rules.

The results will be much worse than insecure computers. Tax loopholes result in less revenue collected by governments, and regulatory loopholes allow the powerful to skirt the rules, both of which have all sorts of social ramifications. And while software vendors can patch their systems in days, it generally takes years for a country to amend its tax code. And that process is political, with lobbyists pressuring legislators not to patch.

It will be very interesting to see how tax and other areas are changed by this sort of thing. Probably turbulent times in the interim though.  

LLMs can be handy for... fixing bad user experiences, or restricted platforms:

https://mastodon.social/@gvwilson/116705903116624751

Via dyne.org, Project Nomad: a local-AI-powered offline knowledge hub "a self-contained, offline survival computer packed with critical tools, knowledge, and AI to keep you informed and empowered—anytime, anywhere."

Pleased to see the Open Knowledge Foundation taking a thoughtful approach to LLMs in a tricky data context: an honest reflection on the integration of LLMs into open data portals. (although I winced at that 'honest'). In Brazil, they are piloting a CKAN-based system that lets people 'talk' to public data. 

Via Tin at the Green Web Foundation -- infratype, CLAUDE: a cloud exchange built by the Journalism Cloud Alliance for investigative newsrooms and public-interest organisations working under real risk, designed to reduce dependence on hyperscalers. 

Douglas Rushkoff used claude to make a website "The internet felt like this in 1994":

when I was done, I held my breath and took a look at how much energy it cost to do all that work. ... I checked the tokens I’ve used so far for my entire Claude experience? Just under five dollars worth.

We could debate the subsidy levels here. Anyway: 

And as far as intellectual property, my AI web designer partner was benefiting off the design and user interface strategies of a myriad of human designers, as was I. So were all the alternative platforms I might have tried. It was getting hard to take an absolutist stance against this tech.

he went to a hackathon:

My favorite project was the simplest: an “I have/I need” bulletin board - basically, the original Craig’s List on digital steroids. An Amazon killer, where people list what they are willing to give away or provide in service, as well as what they need. The kid who needs a bike gets one from the kid who outgrew hers. And because it’s built on AI, it can be a dynamic database that matches people geographically when it’s a thing like a bike, linguistically when it’s verbal, or stylistically when it’s clothes.

Moreover, because it’s being conceived by social activists instead of a team of hired engineers, it’s valuing horizontalism, mutual aid, and trust instead of expedience, profitability, or scale. At worst, they get a working prototype for something they can bring to a non-profit, who can bring on real engineers to build the more durable version. Their working prototype didn’t even cost the five bucks of tokens. 

He's not unaware of the many issues, including AI-powered surveillance. 

Anthropic, the company behind Claude, is supposed to be the good guys—dedicated to human-centered AI and strict guardrails against all this nastiness. And while they’re putting up a pretty good fight against Trump’s efforts to commandeer all AI technology for his crackdown on dissent, they themselves admitted they have to rescind their initial promise not to release AI models if they can’t guarantee proper risk mitigations in advance.

Now on the one hand, it’s a more honest stance. Who can guarantee anything about a technology like AI, which has emergent properties and behaviors no one can really predict? If they spend time and energy on guardrails that may not even work, they will be outpaced by all the companies who don’t give a shit about such things. But if they don’t, then is it a safe place to build the pro-human, pro-social applications my friends and I are conjuring together in the back room of a vegan restaurant?

I like to think Anthropic’s refusal to become part of the US government’s militarization and surveillance apparatus is more important, and a better place to draw their red line. I’m less afraid of a rogue AI than I am a rogue president or a dozen rogue tech billionaires using AI. 

As I see it, the object of the game is to weigh the positive potentials of these technologies against their extractive and sociopathic ones. To treat them like any other technology with dangerous downsides: Is this car trip worth the gas, the pollution, the oil wars? Is this YouTube post worth the algorithms and data servers? Is this vibe-coded program worth the AI cycles? 

Highlights mine. Most of all: 

... Can we lean into the liberating, pro-human capabilities of these technologies before they become unrecognizably incapacitating? ...  Could this moment be different? Did we learn our lessons? Instead of using AI to make cheap, soulless replicas of Hollywood movies and putting creatives out of work, might the real opportunity here be to build platform cooperatives and community-created, worker-owned alternatives to Uber and Amazon?  

Via Luis Villa, Ada Palmer on hopepunk and the futures of hard work. 

Fiction does not give us many stories of continuing to slog on after an unsatisfying partial victory. That makes hopepunk powerful.

... In Hopepunk, people—often ordinary people, including minor characters—take a stand, resist, work together, follow through and help each other, and in the end, while some characters make bad choices, enough make good choices to leave a positive sense of the capacity of humans to choose good.  Put another way, hopepunk presents an image of human beings where, in a prisoner’s dilemma situation, not everyone but enough people actually do choose the thing that helps everyone to make it possible to make the world a better place.  So many stories teach us that, when crisis hits the fan, it won’t take long for biker gangs bedecked with human skulls to rampage through the devastated streets, and very few depict how studies show people really behave in crisis, banding together to supply pop-up pantries and mutual aid. 

... Punk is grungy in aesthetic, and hopepunk shares that, building better among the garbage of the bad.  It also expresses negative emotions, not despair but productive anger, as well as kindness which sometimes needs to take the form of confrontation, or calling someone out.  Hopepunk showcases resilience by showing failure, setbacks, and compromise, not as heroic flaws or formative backstories, but acknowledging that messing up is an unavoidable part of taking action in the first place. 

... Hopepunk narratives are genre stories which have depictions of human nature (teamwork, honesty, resilience) but which also counter purity narratives, by having space for partial victories, unfinished projects, compromise, and mundane not-character-defining failures and mistakes. 

 Here's the original toot I spotted: 

https://social.coop/@luis_in_brief/116564791290630365

 

I think there's something in what Luis says about the open movement and purity culture, but also perhaps something broader about past luxury which also meant things were easy, and so the movement thrived easily, and did not build the toughness or resilience that might serve well in less easy times. (The changing power structures of software developers and tech workplaces of late is quite something.)

I am now reading Alexis Shotwell's Against Purity.

Via John Elkington, a new report from the Impact Economy Foundation. He writes:

In perhaps the most convincing iteration yet of my 1994 concept of the triple bottom line, the report gives the problem a name: the Profitability Gap. Despite the availability of genuinely impactful technologies and business models, companies that lead on sustainability, or are actively transitioning toward it, face a persistent competitive disadvantage.

Markets still reward companies that externalise their costs — shifting the bill for climate damage, biodiversity loss and social harm onto people, nature and future generations — while penalising those that do not.

The numbers are stark. Sixty-six percent of the profits of globally listed companies would disappear if their social and environmental costs were fully accounted for. Current profitability, in other words, depends less on creating value than on avoiding the costs of destroying it.

 Emily Herring considers parallels between gen Z, and the frustrations of young Europeans in the early 19th century:

Feelings of melancholy and ennui were so widespread among Musset’s generation that they were grouped under a single diagnosis: le mal du siècle (literally ‘sickness of the century’).

 danah boyd writes about how social media is the wrong term:

When practitioners used the term “social media” to describe the internet tools that emerged in the mid-aughts, they were giving a name to the kinds of platforms and protocols that allowed people to socialize with friends and communities of interest by using digital technologies. Twenty years later, users of social media are far more likely to scroll than post – and the content that they consume is often strategically produced and algorithmically curated. In this essay, I argue that the very essence of social media has changed. To more effectively interrogate what we are witnessing, we need to stop presuming that these tools are “social media” and begin recognizing that they are now “parasocial media.”

Closing on something generative: The British Cræft Prize, a £60,000 award to fuse heritage craft with cutting-edge technology "not to preserve ashes, but to light a new fire."

We are seeking innovative responses to today’s biggest challenges, inviting inventors to combine the deep wisdom of crafts of the past with cutting-edge technologies of the future.