Fenner Tanswell

Call for Abstracts
The British Society for the Theory of Knowledge invites contributions to its 2026 conference (Exeter College, University of Oxford, 2nd-4th 09 2026).
bstk.org.uk/events.html

Please submit abstracts (500 words, suitable for blind review) to info@bstk.org.uk by April 30 2026.

"the fact that a neologism is sometimes decoupled from its theoretical underpinnings and is used colloquially is a feature, not a bug. [...] This is what it means for a term to enter the lexicon: it takes on a life of its own."
pluralistic.net/2024/10/14/p...

"the fact that a neologism is sometimes decoupled from its theoretical underpinnings and is used colloquially is a feature, not a bug. [...] This is what it means for a term to enter the lexicon: it takes on a life of its own."
pluralistic.net/2024/10/14/p...

To be fair, he is very positive about people broadening the term. He even says it is a good think that there is a technical term that can be applied narrowly, and an intuitive term that can be used broadly, and that both coexist.

Boycotts and Beyond: How should we respond to ethically compromised sporting events? Major sporting events can be ethically charged. They may be hosted by fascist regimes, as a way to gain a perception of legitimacy on the world stage. Notably, the 1936 Olympic games was hosted by Naz...

I'm organising a workshop in Glasgow in May, about how people should respond to ethically tainted sporting events. I've just made a PhilEvents page. If you've got any questions about it, send me an email!
philevents.org/event/show/1...
#PhilSky #PhilosophyOfSport #Ethics #Boycotts

The real theorems are the proofs we made along the way.

Journeys in Mathematical Landscapes: Genius or Craft? We look at how Anglophone mathematicians have, over the last hundred years or so, presented their activities using metaphors of landscape and journey. We contrast romanticised self-presentations of th...

And of course the self-promotion that we have thought about this metaphor a lot!

link.springer.com/chapter/10.1...

The Edge of Mathematics Terence Tao, the legendary mathematician, explains the promise of generative AI.

From here:
www.theatlantic.com/technology/2...

Tao: These problems are like distant locations that you would hike to. And in the past, you would have to go on a journey. You can lay down trail markers that other people could follow, and you could make maps. AI tools are like taking a helicopter to drop you off at the site. You miss all the benefits of the journey itself. You just get right to the destination, which actually was only just a part of the value of solving these problems.

Wake up, honey! A new extension to the "maths as journeying" metaphor just dropped.

I made a philosophy cryptic crossword for our postgrad reading party last week. A few people asked about it, so I shared it with @dailynous.com, and it's now available on the Heap of Links: dailynous.com/wp-content/u...
So if you're a special kind of nerd, check it out!
#PhilSky #CrypticCrossword

Hard agree. This is truly absurd.

Discovered that Gmail abbreviates emails from me to "Fenner St. Tanswell". I didn't realise the process of the pope acknowledging my miracles was complete already. How time flies. Will acknowledge prayers by email, but give me a couple of weeks.

cant.stop.laughing

I've become the thing I hate; written a paper with a nice interdisciplinary-friendly start but with a last section so insider baseball that it has an audience of like 4 people

The Promptware Kill Chain

Attacks against modern generative artificial intelligence (AI) large language models (LLMs) pose a real threat. Yet discussions around these attacks and their potential defenses are dangerously myopic. The dominant narrative focuses on "prompt injection," a set of…

Cory Doctorow shared our new paper about LLMs in Maths today in his daily newsletter, and I'm well chuffed.

pluralistic.net/2026/02/13/k...

Yeah, our university tech also regularly breaks.

When is a sausage not really a sausage? Ask the meat lobby | George Monbiot European legislators may ban plant-based products from using the name to prevent ‘confusion’. Just don’t mention beef tomatoes or buffalo wings, says Guardian columnist George Monbiot

Monbiot did a hilarious take-down of this kind of reasoning.

www.theguardian.com/commentisfre...

New paper! Joint work with @asgeirberg.bsky.social

"The Philosophical Prospects of Large Language
Models in the Future of Mathematics"

Online now in the Annals of Mathematics and Philosophy: mxphi.com/wp-content/u...

Please share it with anyone interested in LLMs and Mathematics!

🧵

Love it, I would read that. :)

As Thurston says? On using quotations from famous mathematicians to make points about philosophy and education - ZDM – Mathematics Education It is commonplace in the educational literature on mathematical practice to argue for a general conclusion from isolated quotations from famous mathematicians. In this paper, we supply a critique of t...

You mean after their discussion of Thurston, an updated one on Tao?

link.springer.com/article/10.1...

hate when reverse centaur keeps happening

I think that he has changed directions a bit since we wrote this, focusing a lot more on the interactive version. Some but not all the same argument can be used there.

Go read our paper!

Please do share this with anyone you think might be interested!

There is plenty more to say. E.g. what if the LLM is used interactively? We have lots to say about that, but that is not in this paper!

Relatedly, there is the classic question of whether this is all still proper mathematical justification or merely probabilistic. Trusting a proof because an autoformalised version passed a Lean check might be highly reliable, but it sounds a lot more like a probabilistic justification.

Okay, then check that it doesn't do that. But checking code is hard. Oh no, you're a reverse centaur again. Better keep vigilant for a typo that messes up hundreds of lines of code!

The next answer would be: then formalise the proof and check it with a computer. Our paper discusses autoformalisation a bit, and how the dangers are still present. For example, an LLM might formalise a proof to something true but trivial, rather than the intended theorem.

“[…] in many cases, the solution the chatbots provide is very long, detailed, and written in a “professional” way, but it still may be completely wrong, or make no sense at all when examined more carefully. This may fool a human to think that such a detailed and long solution would be correct, so extra caution is needed when we use such tools for solving similar exercises.” (Emphasis ours. Plevris et al. 2023, p. 18)

Human mathematicians do have experience and expertise spotting human errors, but LLMs do not fit the same patterns, so their purported proofs are harder to check. They are not following norms, merely imitating them. They lack common sense. They make weird and unpredictable mistakes.