Thought Shrapnel

The widely-referenced “stochastic parrots” paper from five years ago is no out of date. In it, Emily Bender, Timnit Gebru, et al. argue that LLMs remix patterns in text without genuine understanding. This has knock‑on effects for how we (should) use and trust them. It’s a familiar argument, using the same approach as John Searle’s famous Chinese Room argument about ‘black box’ symbol‑shuffling without understanding.

I don’t know Pete Wolfendale, but I have just discovered that he is an independent philosopher based in Newcastle‑upon‑Tyne, so I should probably look him up. In this post he pushes back on the idea that LLMs are “stochastic parrots” and uses that as a way into some bigger questions about what we mean by “meaning” in the first place.

I think it’s maybe worth summarising why I think post-Searlean critics of AI such as Emily Bender are wrong to dismiss the outputs of LLMs as meaningless. Though it’s perhaps best to begin with a basic discussion of what’s at stake in these debates.

Much like the term ‘consciousness’, the term ‘meaning’ often plays proxy for other ideas. For example, saying systems can’t be conscious is often a way of saying they’ll never display certain degrees of intelligence or agency, without addressing the underlying capacities.

Similarly, being able to simply say ‘but they don’t even know what they’re saying’ is a simple way to foreclose further debate about the communicative and reasoning capacities of LLMs without having to pick apart the lower level processes underpinning communication and reasoning.

While Wolfenden agrees that current systems don’t have beliefs, intentions, or a stable view of the world, nor does he think that they’re “just” meaningless text generators:

1. Grounding - a lot of human language isn’t directly tied to what we personally see or do. We can talk meaningfully about things like black holes, stock markets, or bowel cancer because we trust expert communities - not because we all have first‑hand access. So it’s at least plausible that LLMs pick up some real, socially grounded content from the human language they’re compressing.
2. Intention - while he accepts that LLMs do not have inner communicative goals, Wolfenden suggests that this is not as much of a big deal as critics suggest. He compares the outputs of LLMs to rumours - i.e. statements produced by a diffuse social process. We can still interpret, question, and trace this back to wider patterns of usage. LLMs sit inside our language community in this way, even if they are not full participants.
3. All-or-nothing thinking - Wolfenden argues that we should avoid binary thinking about minds and meaning. Humans also talk lazily, gossip, and say things we later revise. We can be asked for reasons for our views and have our views be reshaped, which is the difference between us and current models. There is still overlap here. As he says, LLMs are “in the game, even if they’re not strictly playing it.”

Source: DEONTOLOGISTICS

This post by Dave Snowden, originator of the Cynefin framework, relates to post I shared by Tom Watson about ‘loose ends’.

I have been writing recently, and will write more, about the difference between containers and landscapes: how interventions can be real within their boundaries and yet leave everything structurally unchanged outside them. The coaching session that produces genuine insight. The workshop that shifts something in the room. And two weeks later, nothing. The container was real. The landscape was unchanged.

[…]

I have written elsewhere about what a different approach might look like in practice: working obliquely, creating conditions for empathy through shared action rather than mediated dialogue, allowing conversation to arise from mutual work rather than being engineered by a third party with an agenda. The Derry Girls scene captures the failure mode with comic precision. The reconciliation workshop produces nothing. The two smiles exchanged across the room, unprompted, unmediated, while chaos erupts around the parents, are where something real briefly appears. You cannot programme that moment. You can only create the conditions that make it possible.

Source: The Cynefin Co.

Everything that’s wrong with the world, captured neatly in one cartoon.

Source: Tom Gauld

Tom was talking to me about his thinking about this post when we met up earlier this week to discuss next steps for TechFreedom, our joint project.

Essentially, the problem is that things like workshops, events, projects, and even programmes of work have an internal logic to them. This logic dictates whether or not they are designated ‘successful’. Whereas, the world is a messy and complicated place, and simply giving people opportunities to connect and think things through can have much more profound consequences.

Most of what happens in a workshop or a session is only ever useful if people can take it back into their own context and make sense of it there. A good facilitator will create rooms and spaces that support the workshop. But however well designed, the room is artificial. The real work happens when someone is back at their desk on a Tuesday morning trying to figure out what any of it means for the decision they’re actually facing.

If we resolve everything in the session with neat actions, clear conclusions, a satisfying arc, there is the potential we’ve artificially done the sense-making for them. We’ve removed the productive friction of trying to figure out what it means to me in my context outside the room. We’ve made it easy to file the experience away.

But what if we leave a question hanging, one that is genuinely & intentionally unresolved, not because we ran out of time but because we chose to? One that forces them to contextualise, to test an idea against their own reality, to keep thinking after the room has emptied.

Maybe the loose end isn’t a failure of facilitation. In some cases maybe it’s the mechanism.

Source: Tom Watson

I’m not saying that you need to be an expert on the history of every country of the world, but when there’s a major crisis going on, understanding why it’s all kicking off is at least worth understanding.

Source: Ted Rall

Source: Are.na

It’s almost a decade since one of the greatest economic harms a country has ever self-inflicted. Yes, I’m talking about Brexit.

Finally, we’re getting to the stage when our current government, which is not the one that instigated the referendum, can say that “Brexit did deep damage”. Let’s hope we get back into bed with our European neighbours ASAP. The decline in Britain over the last 10 years is tangible.

“Brexit did deep damage.” With those words at her Mais lecture on Tuesday, Chancellor Rachel Reeves made it clear that there has been an important shift within the Labour Party - one that government ministers have been signalling for some time.

“Let me say this directly to our friends and allies in Europe. This government believes a deeper relationship is in the interest of the whole of Europe,” she said, while at the same time insisting that the government was not trying to “turn back the clock” on Brexit.

Speaking in such overt terms about Brexit’s perceived harms in part reflects a belief that, as the government attempts to turn around the country’s persistently sluggish economic performance, it must be more ambitious in its attempt to “reset” the UK’s post-Brexit relationship with the EU.

[…]

Speaking at a literary festival in October, Wes Streeting, the health secretary, said: “I’m glad that Brexit is a problem whose name we now dare speak,” and indicated that he believed being outside the EU was making it difficult to deliver the economic growth the government had promised.

The deputy prime minister, David Lammy, said in a podcast that it was “self-evident” that Brexit had damaged the economy and noted the economic benefit that Turkey had derived from its customs agreement with the EU.

Meanwhile, in further evidence of pressure within Labour’s ranks to rethink its policy on Brexit, on Wednesday the London Mayor, Sadiq Khan, called for the UK to rejoin the EU customs union and single market before the next election, and then campaign at that ballot on a promise to rejoin the EU.

Source: BBC News

This is definitely not for everyone, but ContraPoints (Natalie Wynn) is awesome and always makes compelling videos. This one weighs in at a little over an hour and a half, so I’m still watching it.

Ever since studying Philosophy of Art & Literature as an undergraduate, in which we looked at why people watch horror films, I’ve understood that disgust is actually a complicated emotion. As ContraPoints explains through the Saw series of films, so-called film critics have things all wrong.

I love the provocation that Home Alone is a a more ethically problematic film that Saw because we identify with the aggressor (Kevin) rather than the victims (the Wet Bandits) in the former.

The closest I get to horror these days (or any day) is watching films like Sinners which I watched with my son at the cinema recently. Once you understand that there is an art to these things, and that they have lessons for us as humans, it opens up a whole new world.

Source: YouTube

I did not enjoy studying Formal Logic as a Philosophy undergraduate. But it stood me in good stead.

I’m pretty sure there are plenty of people who wouldn’t even understand what’s wrong with the above reasoning, and in fact it explains a lot of what is wrong with the world… 🙄

Source: X via Are.na

Yes, Claude Cowork is great, but the secret sauce is actually Claude Code which you can access via the Claude app. Even better is doing so from the command line interface (CLI).

The advantage of the CLI is that you’re fully in control of your project. The difficulty, of course, is that unless you grew up having to load computer games via DOS, and unless you’ve got a mental model of how product development works, it’s going to feel very odd.

Source: How to AI

Whatever you call it, having everything in space you control has always made sense.

Why am I doing all this? Because I got inspired by the concept of POSSE: “Publish on your own, syndicate elsewhere.” For me, ROOTS is the logical first step toward that: “Return Old Online Things to your own Site” (yes, I made this up). Why? If I do decide to delete my X account or if Blogger gets quietly discontinued, then I don’t care: it’s all on my site already. I own it. It’s all Markdown files and images that I can back up anywhere I want.

You’ll see me POSSE (or PESOS – “Publish Elsewhere, Syndicate (to your) Own Site”) in the future, too: If I post a Goodreads review, it’ll also be on my Notes and Everything pages. If I post on LinkedIn, it’ll be there, too. Everything I create and find important will eventually end up on my website.

Source: Lisa Charlotte Muth

Image: GG

An insightful post from Max Read about stylistic preferences with regards to human vs AI text. Every relevant technology changes writing and, in turn, literate culture.

In many contexts most people can (more or less) correctly differentiate between A.I.-generated output and its “authentic” counterpart–but cannot correctly attribute the output.

What’s funny about this is: We actually really want to prefer human-authored writing! In open-label tests, where the excerpts are shown with attribution, people consistently express preference for whatever text is labeled human, even when the text is actually A.I.-generated. (So do A.I. evaluators, as I learned at the conference from Wouter Haverals, to an even greater degree.)

This is not a particularly satisfying set of findings insofar as it validates neither the A.I.-booster “it’s so over, A.I. writing is better than human writing” side nor the A.I.-skeptic “A.I. can never write like a human” side. What we can say is that people mostly can’t identify A.I.-generated text as A.I.-generated (crowd boos), but they can sometimes distinguish between it and human-authored text (crowd cheers); it’s just that they tend to think the A.I.-generated text is human (crowd boos), maybe because human-generated text is stranger, worse, or more difficult (crowd hesitantly cheers), which readers mistakenly believe is more typical of A.I.-generated text (crowd silent now) and thereby disprefer (crowd sort of murmuring confusedly), unless you tell them it’s actually human, in which case they change their minds and like it (crowd has mostly left at this point).

But all of it taken together suggests that, given our strong bias in favor of writing we believe to be human, A.I. vs. human “preference” tests (or “reads better” quizzes) are often second-order “identification” tests, in each case measuring not “preference” per se but the accuracy of the prevailing heuristics for identifying A.I. writing. Participants in these studies, it would seem, express preference for the A.I.-generated writing not because it’s “better” in some formal sense–cleaner, simpler, more beautiful, whatever–but because their “flawed heuristics” have led them to the conclusion that it’s human-authored, and ipso facto better.

[…]

As long as people want to prefer human-authored to L.L.M.-generated writing, we will place a premium on whatever style we associate with human authorship–even as that style changes. You can already see this process beginning from the other direction on social networks like Twitter, where em-dashes and not-x-but-y contrastive corrections–perfectly innocuous and useful writerly tools which not five years ago would likely have been highly correlated with “good prose”–are immediately treated with derision and suspicion. By that same token, certain kinds of “bad writing” should be seen as evidence of human authorship. How long before run-on sentences are preferred to em-dashes?

L.L.M.s, of course, can and will get better at mimicking the “strangeness,” clunkiness, and badness of human prose; I’m skeptical of claims that there is some built-in technical limitation that prevents A.I. text from ever being truly indistinguishable from human prose. What seems more likely to me is that as L.L.M.s move away from the easily identifiable generic LinkedIn style that currently dominates, our preferences will move as well, in an attempt to stay one step ahead.

Source: Read Max

Image: Randy Tarampi

Source: YouTube

Five years old, but still as relevant as ever.

Source: Bill Bramhall

Link to video

I ran a 90-minute workshop this morning, which started life as a 1:1 session. Around 10 people ended up coming, mainly from a couple of Slack channels - hence the “Hey Slackers…” intro.

Below is the email I sent afterwards with all of the links, etc. I’m posting it here for reference 😀

Thanks for joining this morning’s session, or for registering if you couldn’t make it. Here’s a recap of what we covered, plus the resources to help you get started.

What we did

We looked at how freelancers can use AI tools like Claude to build a business-development dashboard — a lightweight CRM and pipeline tracker that works for your specific situation, not some generic corporate workflow.

The conversation covered a lot of ground, with great contributions from everyone:

• Separating your worlds — several people mentioned mixing client work, personal projects, and other commitments (one person had a part-time Masters degree in the mix) inside a single to-do list. A dedicated dashboard draws those lines clearly.
• Tracking invoices and payments alongside pipeline activity — this came up as a popular request.
• The “second brain” approach — capturing and evaluating ideas, not just managing active leads.
• AI as a conversational partner — the key shift is thinking of it like onboarding a business development colleague who learns your context over time, rather than expecting a one-shot answer.
• Environmental and ethical considerations — there was a thoughtful discussion about AI scepticism, resource usage, and how local LLMs work well for lighter tasks (restructuring text, for example), with remote tools better suited to heavier ones (like web scraping).
• Keeping a dashboard updated — the question of maintaining a CRM built from an existing contacts export was live in the room. Cross-referencing with LinkedIn was one practical starting point.

Meta prompt The only really thing you need to remember is to give AI tools permission to ask you questions. One ‘meta’ prompt that I’ve come across which can be useful is:

Don’t answer my question yet.

First do this:

1. Tell me what assumptions I’m making that I haven’t stated out loud
2. Tell me what information would significantly change your answer if you had it
3. Tell me the most common mistake people make when asking you this type of question Then ask me the one question that would make your answer actually useful for my specific situation rather than anyone who might ask this.

Only after I answer — give me the output

My question: [paste anything here]

Tools mentioned

• Claude / Claude Cowork
• ChatGPT
• Google AI Studio
• Mistral AI Studio
• Monica HQ — open-source personal CRM
• Groundwork and TaskDial — Doug’s own projects
• Zen Browser — a Firefox-based browser worth a look

A note on AI and ethics

One participant raised points worth sitting with: hallucinations, the concentration of money and influence in big tech, resource usage, and what we lose by outsourcing too much thinking. These are fair concerns. Start small, stay critical, and use local tools where they’re up to the job.

More reason to divest yourself of US-based Big Tech platforms. (Join the first TechFreedom cohort!)

Iranian state-linked media this week published a list of offices and infrastructure run by American companies with Israeli links whose technology has been used for military applications. According to Al Jazeera, the companies include Google, Microsoft, Palantir, IBM, Nvidia and Oracle.

Many of these companies operate regional offices, cloud infrastructure or data-centre operations across the Gulf, including in the UAE. None have released public statements on this development.

The list was published by the semi-official, Islamic Revolutionary Guard Corps-linked Tasnim News Agency alongside a warning that the scope of the conflict could expand beyond traditional military targets.

“As the scope of the regional war expands to infrastructure war, the scope of Iran’s legitimate targets expands,” Tasnim News Agency reported.

Last week, Iranian drone strikes damaged Amazon Web Services (AWS) data centres in the UAE and Bahrain, disrupting services and exposing the vulnerability of physical tech infrastructure in the region.

Source: WIRED

Image: Mockup Free

I’ve always been against unearned privilege and the idea of a ‘natural’ hierarchy. It’s antithetical to who I am and stand for, and I’ve felt that way ever since I can remember.

A good example of this in the class-stratified UK is the House of Lords. While it’s important to have a second chamber in a democracy, the idea of it being made up of heredity peers is absolutely ridiculous.

So I’m delighted that Labour are finally getting rid of the absurd idea that, just because someone is descended from an ancestor who was given land by William the Conqueror, they should have a say over our democratic processes.

Next? How about we rename the highest honours civilians can attain, removing the word ‘empire’ from OBE, CBE, etc.

Centuries of British political tradition will end within weeks after Parliament voted to remove hereditary aristocrats from the unelected House of Lords.

On Tuesday night members of the upper chamber dropped objections to legislation passed by the House of Commons ousting dozens of dukes, earls and viscounts who inherited seats in Parliament along with their aristocratic titles.

Government minister Nick Thomas-Symonds said the change put an end to “an archaic and undemocratic principle.”

“Our parliament should always be a place where talents are recognized and merit counts,” he said. “It should never be a gallery of old boys’ networks, nor a place where titles, many of which were handed out centuries ago, hold power over the will of the people.”

The House of Lords plays an important role in Britain’s parliamentary democracy, scrutinizing legislation passed by the elected House of Commons. But critics have long argued that it is unwieldy and undemocratic.

Source: Associated Press

Image: João Marcelo Martins

This made me laugh. Entitled Tree Hug by Bulgarian street artist Vanyu Krastev.

Source: Street Art Utopia

Now That’s What I Call An MVP

Source: News Canary

An inability to focus is a design problem. As I noted back in 2020 on my now-defunct literaci.es blog, perhaps we need notification literacy (archive.org link).

If the problem is screens inherently, then we need cultural revival, a return to books, perhaps even a neo-Luddite retreat from technology. But if the problem is design, then we need design activism and regulatory intervention. The same screens that fragment attention can support it. The same technologies that extract human attention can cultivate it. The question is who designs them, for what purposes, and under what constraints.

In the library, I watch people navigate information in ways that would have seemed impossible to previous generations. A research question that once required weeks of archival work now takes hours. But more than efficiency has changed. The nature of synthesis itself has transformed.

Ideas now move through multiple channels simultaneously. A documentary provides emotional resonance and visual evidence. Its transcript enables the precision needed to locate a specific argument. A newsletter unpacks the implications. A podcast allows the ideas to marinate during a commute. Each mode contributes something the others cannot. This isn’t decline. It’s expansion.

What strikes me most is the difference between people who’ve learned to construct what I call ‘containers for attention’ – bounded spaces and practices where different modes of engagement become possible – and those who haven’t. The distinction isn’t about intelligence or discipline. It’s about environmental architecture. Some people have learned to watch documentaries with a notebook, listen to podcasts during walks when their minds can wander productively, read physical books in deliberately quiet spaces with phones left behind. They’re not rejecting technology. They’re choreographing it.

Others are drowning, attempting sustained thought in environments engineered to prevent it. They sit with laptops open, seven tabs competing for attention, notifications sliding in from three different apps, phones vibrating every few minutes. They’re trying to read serious material while fighting a losing battle against behavioural psychology weaponised at scale. They believe their inability to focus is a personal failure rather than a design problem. They don’t realise they’re trying to think in a space optimised to prevent thinking.

This is where my understanding of literacy has fundamentally shifted. I used to believe, as I was taught, that literacy was primarily about decoding text. But watching how people actually learn and think has convinced me that literacy is about something deeper: the capacity to construct and navigate environments where understanding becomes possible.

Source: Aeon

Image: Mario Aziz