Bagatelles#
These are papers I wrote over the last few years, some in Oxford (OUSSA), some in Paris, some at home and all for my own pleasure. Enjoy, agree or disagree! Why ‘Bagatelles’? It works in three languages and shows that I don’t take myself too seriously. Why English? It is the Latin of our days, the lingua franca of the Western intellectual world. (You can ask your browser to use a different language.)
No peer reviewers, no limits on the word count, no deadlines. Oh century, oh science, it is a joy to live! (Ulrich von Säckingen, 1518)
I am a committed, shameless, and unrepentant user of AI, and I feel like a farmer in the 1950s, trying out his first harvester. Here is my claim: AI is not a replacement for thought but an accelerator of it. Thanks to Claude Code, this website was put together in just a few hours, starting from a broken, plain-vanilla site and a heap of Markdown files. The introduction to First Steps in Arithmetic was generated by GPT. I could have written a similar text, but not in two seconds. The proofs are mine. Medieval Philosophy * is a summary of a conversation with Claude. It condenses my view of the subject and is useful when read alongside other sources such as [Russell, 2004] and [Adamson, 2014]. In Problem of Free Will Dissolved * I defend my view against Claude’s objections, as I do in What Exists in Mathematics? *. If you ask Claude to challenge you, it will probably give you more than you asked for. I need a rest after two hours of objection–response exchange, but I accomplish in days what would otherwise have taken weeks or months.
The papers marked with * are AI-assisted.
They express exactly what I want to say — often better than I could have done it myself.
I am not the author in the traditional sense; rather, I am the Executive Director.
I can sign the paper and stand by its content, exactly as did the boss in the old days,
whose other collaborators didn’t appear on the title page. I could also conceal the fact that AI was used:
after some reworking, possibly with the help of AI, I defy any expert or AI tool to tell the difference.
In fact, you could even train an AI to write in the same funny, old-fashioned style of Einstein, Heisenberg and others!
Is this a challenge for editors and publishers? Does AI assistance devalue an otherwise excellent and informative paper or discovery?
I can’t see why it should. And what about the credit to the authors? This problem has existed since the beginning of
scientific publications.
The famous EPR paper [Einstein et al., 1935] was written by Podolsky,
and Einstein apparently did not see (certainly did not inspect) the final draft before submission, see [Fine, 1996], p. 35.
As an outsider you just don’t know how the contributions of the various authors were distributed.
The student who copy-pastes his subject into GPT and forwards the
generated text to the tutor is an extreme case. It is a caricature and a challenge for the universities,
but no argument against AI.
But there are serious arguments: the intellectual-property issue (a challenge for legislators), the energy footprint (but many AI providers are committed to renewable sources), the job market (how can the value added by AI be distributed fairly?), the spectre of singularity (no evidence but no way to refute it) — to name just a few. Despite these concerns, AI can contribute greatly to solving our problems and making life more worthwhile.
Yet it can also be misused to lead the world into disaster. [O'Neil, 2016] gives hair-raising examples of the havoc that algorithms can wreak. Social media, with their horrible impact on opinion-making, are assisted by algorithms and AI. And here is the towering threat: Those who hold the power, the ones in office, are they good guys or bad guys? I am not too optimistic, given the large number of voters who seem to have lost their marbles. Or have I lost mine? You decide.
Johannes Siedersleben