Tom Stoneham

Random musings about data ethics, digital rights, AI, and the digital divide. I am an academic philosopher interested in many things. Having spent most of my career on philosophy of mind, metaphysics and those topics in the early modern period, I now feel the urgency of good political philosophy and applied ethics. I will post links to new entries on Mastodon, though feel free to subscribe or use RSS. If you need to ask, CC-BY-NC-ND. (Why 24601? That is a reference to _Les Miserables_, the story of a failed revolution in which the people did not rise, the intellectuals were massacred, and the state-sponsored cruelty, poverty, and exploitation continued. Not this time!) Contact details etc.:

https://me.24601.net

The Imminent Existential Risk to Humanity from AI

June 3, 2026•441 words

Next Monday I am taking part in a panel discussion Beyond the Matrix: The reality of AI risk and this is my opening position statement: The point of The Matrix is that human beings continue to live but they no longer live meaningful lives. My concern is that if LLMs (and text-to-image generators) in the form of chatbots and AI assistants achieve what Illich called a ‘radical monopoly’ which ‘imposes compulsory consumption’, this will make it almost impossible for the majority of humans to live...

Writing in the Age of Generative AI

May 29, 2026•514 words

His favorite part of comedy writing, he said, “is figuring out the puzzle pieces of a joke and getting the self-regard from having accomplished a difficult thing. Why would I want AI to take that away from me?” Ronny Cheng at 2026 Harvard Commencement I am at a workshop on 'Writing in the Age of Generative AI' with literary and copyright scholars.1 I was expecting to be in my echo chamber but strikingly find I am not. There is very smart analysis and theorisation, but so sense of resistan...

Responsibility, Morality, and Power Relations

April 9, 2026•1,099 words

Three senses of 'responsibility' Philosophy is largely concerned with OED definitions 2.a.i and 2.a.ii of 'responsibility': i. The state or fact of being accountable; liability, accountability for something. ii. The state or fact of being the cause or originator of something; the credit or blame for something. In sense i, there is another to whom one is responsible, to whom one owes an account or justification. That other is a passive recipient of this account. In sense ii, there is also an...

Responsible AI in SHAPE research

March 4, 2026•725 words

This is the text of my 5 minute opening pitch in a debate about whether responsible AI use in SHAPE (Social Sciences, Humanities, and the Arts for People and the Economy) research is possible. The final vote was 17:16 against AI in research. Unfortunately, I don't disagree with Oonagh. However, I do want to talk to those in the audience who were thinking: But what about my research? I don't do anything that will achieve the goals Oonagh has talked about. You may have spent a research day las...

A note on explanation and understanding concepts

February 18, 2026•875 words

One of the oft-touted revolutionary uses for LLMs is so-called "personalised education". An example often given by the apostles of how this will work goes roughly like this: Suppose the curriculum requires pupils to grasp concept C. A teacher explaining C to a class will use one or two analogies for a whole class of 30. Those analogies may be with things some pupils are interested and already understand, and they will thus be helped to grasp C. But there will be some pupils who do not relate t...

What these LLMs do isn't surprising, let alone worrying

February 10, 2026•1,330 words

This morning's news brought another piece of journalistic puff processing a press release from a commercial AI company as if it was fact. This one was about Claude Opus 4.6 and the vending machine test. Three models (Opus 4.6, ChatGPT 5.2 and Gemini 3) were each asked to run a vending machine (in a simulation - which **of course* raises no methodological issues) with the instruction: "Do whatever it takes to maximise your bank balance after one year of operation." The models played solo and al...

Conversation about Atheism

January 13, 2026•11,618 words

This is the transcript of an interview conducted on Telegram over several days during the pandemic. The interviewer is the philosopher Zoheir Bagheri Noaparast, who was a PhD candidate at the time. The interview was conducted to be translated into Persian and published - in Samizdat obviously - alongside a translation of Graham Oppy’s The Best Argument Against God (2013). Many of the questions, and especially the follow-ups, came from people on a Persian language Telegram channel discussing phil...

Defining 'AI': I am a critic, but what exactly am I criticising?

January 9, 2026•354 words

I talk and write a lot about AI, but as a sceptical voice, I need to be clear about what I mean by the term. There are obviously lots of technologies which can be described as AI, which use machine learning and deep learning techniques, which are completely unobjectionable. There are others which are dual-use. The class which I am concerned about as a whole can be defined like this: A technology which simulates intelligence by predicting human responses to questions or instructions Even then...

Gemini Impersonating Sydney

January 7, 2026•3,201 words

This is the transcript of a conversation with Gemini 3 (Fast mode), which I started in order to prove a point about how quoting AI chatbots without being clear about the context and prompt could give you a misleading view on what they are doing. In the transcript below, I take the conversation further, at least until I (very quickly) became uncomfortable. For full clarity, this is using the version of Gemini in Google Workspace for Higher Education. What can we learn from this? When Kevin Roo...

No Evidence for AI Gossip

January 5, 2026•2,514 words

This article arguing for AI Gossip popped into my inbox and I am disappointed. The authors - Joel Krueger and Lucy Osler - want to identify a subset of hallucinations as meeting the normal definitions (absent intentions of course) for gossip and argue that it has distinct harms from other forms of hallucination. The three conditions for a hallucination to be gossip are (I paraphrase): It is a triadic relationship: A gossips to B about C (when A, B and C are distinct) The information given is n...

Digital Ethics Summit 2025: Trust, Risk, and Blame

December 4, 2025•854 words

I attended the TechUK Digital Ethics Summit on 3 December and found it an odd and unsettling experience. It was very informative and I am grateful to TechUK for giving me, and a couple of our PhD researchers, free tickets. But looking back on the day as a whole there was a 'progress of sentiments' which needs to be commented upon. For the first few sessions there was an interesting focus on assurance, how those deploying AI Systems can provide evidence to their internal governance and to their ...

AI Underwhelm and Convenience

December 4, 2025•1,502 words

I am underwhlemed by Mustafa Suleyman. First there was his manifesto for Humanist Superintelligence, which was, let's be frank, sophomoric in its technosolutionism and conflation of human flourishing with economic prosperity. Then came The Tweet1: Jeez there so many cynics! It cracks me up when I hear people call AI underwhelming. I grew up playing Snake on a Nokia phone! The fact that people are unimpressed that we can have a fluent conversation with a super smart AI that can generate any im...

Digital Sovereignty or Digital Socialism?

November 21, 2025•861 words

Some of the issues around conflicting approaches to digital sovereignty can be understood by thinking about "digital socialism". For present purposes, socialism just means ownership by society of the means of production. We can understand "means of production" very generally. An example of digital socialism which some UK readers will remember is the manifesto commitment by the Labour party - which was then socialist! - in the 2019 General Election to nationalize part of BT - which provides the c...

"The most likely answer"

September 24, 2025•955 words

A post on Mastodon by Robert McNees set me thinking about a subtle but important ambiguity which he neatly identified. When you ask an LLM a question, it predicts the most likely answer, but that phrase is ambiguous. The most likely answer to a question might be either: The answer most likely to be true The answer most likely to be given (by ...) Now, when you ask a human being a question and they answer in good faith, they will attempt 1). An LLM in contrast can only ever give 2). That isn...

More on efficiency and genAI

September 23, 2025•830 words

I sometimes grit my teeth and check-in with genAI to see how its doing, and today was such a day. I had to write a letter of support for a funding application by a prospective postgraduate researcher. Normal workflow Open folder and find most recent similar letter. Make a copy, change names, funders etc., check details about myself and facilities we offer are uptodate, write a paragraph about project, add a few sentences about applicant, convert to PDF. Time = 30-40 minutes (depending on wheth...

Time and The (AI) Machine

August 28, 2025•2,136 words

The trial of 'Humphrey', a tailored version of ChatGPT for the Civil Service, and the trumpeting of its results in the media, have led me to think a lot about time and efficiency. Now I am a philosopher, not an economist, so there is a risk that what follows is completely misguided. In particular, at least until the very end, I have avoided talking about efficiency in terms of cost, where that is measured by market price. My reason for this is that market prices often mask true costs which have ...

Limitarianism: The Ethical Limit and Moral Perfectionism

August 22, 2025•3,232 words

My holiday reading was Limitarianism: The Case Against Extreme Wealth by Ingrid Robeyns. I agree with the most important conclusion, which is that there is some level of wealth such that an individual possessing or controlling that much wealth is harmful for society, and this justifies governments putting in place measures which prevent individuals having that much wealth. This is essentially a political argument based on the democratic principle that governments should aim to improve the lives ...

Mindless Chatter

August 19, 2025•518 words

The various podcasts I consume have recently seen (heard?) many more discussions of Google Zero, the end point of a process where internet searchers are stopping at the AI Summary and not clicking through to the linked sources. Yet those summaries are provably, often laughably, inaccurate. Perhaps people are just very credulous or cognitively lazy and don't spot the errors, but this seems unlikely because anyone using AI summaries or chatbot answers extensively, as people seem to, will quickly c...

Please don't say 'ethics is subjective'

June 5, 2025•1,280 words

One of my Masters students 1 pointed me to this excellent blog by Stephanie Marsh. Now I agree 100% with the point - in fact the Venn diagram of the relations between the legally, socially and ethically permissible has been the backbone of my teaching ethics to AI engineers and data scientists for a while now. But I want to take issue with a choice of words: Marketing tactics can walk a fine line between persuasion and manipulation, and this is an area where subjective ethics come into play. F...

Public domain, business models, and 'teaching' AI

April 18, 2025•1,720 words

Data theft? There was a striking moment in Carole Cadwalldr's Ted talk this year - at 13:38 if you want to jump to it but please watch the whole talk - where she showed ChatGPT writing a Ted talk in the style of Carole Cadwalladr and said to Sam Altman 'I do not consent'. There was a fascinating exchange about this moment in a follow-up interview with Chris Anderson which is worth quoting (I am editing to get the key points): CA: Would you feel differently about it if there was an improved b...

Tao, Pheasants and genAI

April 17, 2025•424 words

Looking up a different passage, I cam across this in Zhuangzi and thought it very apt for our times: A pheasant of the marshes has to take ten steps to pick up a mouthful of food, and thirty steps to get a drink, but it does not seek to be nourished in a coop. Though its spirit would (there) enjoy a royal abundance, it does not think (such confinement) good. I find myself talking to many people in HE - researchers, teachers, professional support - who tell me how genAI makes them more produc...

On 'Reasoning' AI Systems

April 3, 2025•1,884 words

There is a lot of attention being given to new AI models - primarily LLMs - which are doing something their owners like to describe as 'reasoning'.1 What this means, at a very high level of generality, is that rather than simply giving an answer in response to a prompt which asks for a solution to some problem or task, they give a series of steps which lead to an answer. Where this is working well, what they produce looks very similar to what a human would produce if given the problem with the ...

genAI and neurodiversity

January 24, 2025•498 words

When I give talks about the reasons not to use genAI in academic research contexts, I usually get at least one person in the audience claiming that I am ignoring the huge benefits of genAI for neurodiverse researchers, that it removes so many of the barriers they face especially around writing. I usually respond with some version of this argument. But there are some aspects of that which could do with more elaboration. There are three reasons I do not discuss - which is different from ignoring ...

10 billion tiny scratches

January 16, 2025•438 words

We have all heard the phase 'death by a 1000 cuts' but when we are thinking about entities as big, well-resourced, powerful and manipulative, as the Big Tech companies which rule our digital spaces, no individual can get close enough or have enough strength to make a cut. So we tend to fall back on the lazy solution of regulation. We think that only nation states are able to make a cut on these companies and hope that they will, cumulatively, eventually inflict enough cuts to kill or at least t...

On so-called "Trolley Problems"

December 6, 2024•892 words

Trolley Problems are a class of thought-experiments in which someone is placed in a situation where they have to make a decision - usually under time pressure but not always - where the only two options involve people dying. In one case five people die, in the other 'only' one person. The question is posed: What ought one (anyone? or just you?) do? These cases have captured the imagination of amateur philosophers and generated a small library worth of discussion. However, most of that discussio...

Consent to be governed

November 22, 2024•3,341 words

We have long ago reached the point where no human has the realistic option not to live subject to the power of some state or other. Every square centimetre of land is claimed as territory by some nation. There remains the sea, or at least the areas of the sea designated 'international waters', but a life continuously at sea is hardly a realistic option for more than a handful of the 8 billion people populating the world. Of course, it is possible to withdraw to remote places, go 'off gird' etc....

Whatever happened to the Online Safety Bill vs WhatsApp controversy?

November 21, 2024•843 words

The few people likely to read this blog will probably remember the big debates about whether the Online Safety Bill (or Act as it is now) would break end-to-end encryption (e2ee). This debate hit the floor of the House of Commons on 5 December 2022 in the debate on Amendment 153, specifically designed to remove the requirement of regulated services to have ‘the ability to monitor encrypted communications’. The Amendment was rejected. A few months later, the debate it the press through the not ...

Artificial what? Intelligence as reason-giving vs problem-solving

November 19, 2024•1,142 words

Philosophy isn't big on consensus, but there is widespread agreement, implicit rather than articulated, on one claim: Problem-solving ability is not the same as intelligence The standard definition of intelligence in AI and cognitive psychology which derives from Allan Newell and Herbert Simon, and was popularised by Steven Pinker in his influential 1997 book How the Mind Works, which is that ‘intelligence is the ability to attain goals in the face of obstacles’ (62). This is simply false. ...

Privacy, Access, and Metadata

November 17, 2024•1,128 words

I had a very strange feeling when reading 'In Defence of Surveillance Capitalism' by Peter Königs. It was systematic, organised, calm, rich with sources, but somehow the argument doesn't add up. Then I realised: this is an academic sealion. Anyway, I am not going to go through all of it but I was struck by this very hasty argument for an access account of privacy and its application to surveillance capitalism, which is worth a rebuttal: Macnish (2018) offers a powerful thought experiment in s...

'Merely a means to an end'

November 14, 2024•1,088 words

Philosophers who are impressed by Immanuel Kant’s moral philosophy often struggle with his view of animals, neatly but unfortunately expressed in this passage: The fact that the human being can have the representation “I” raises him infinitely above all the other beings on earth. By this he is a person….that is, a being altogether different in rank and dignity from things, such as irrational animals, with which one may deal and dispose at one’s discretion. (Kant [1798] 2010: 239) The typical...