How’s the writing going Ian?

Surprisingly well … despite relatively few productive days amidst the frustrations of everyday life. The deliverables are ever clearer – a PhD-style Thesis and 2 books – and first a paper for ISSS2025 in Birmingham in a couple of weeks.

Mentioned in the previous post – to which I added spin-off footnotes already – that I’d been getting the feeling that every new thing I drafted was something I’m sure I’d already drafted better somewhere before? So, I pointed ChatGPT at a selection of my own existing writing (linked in that post) and asked it to summarise it all back to me.

That’s become the start of a little adventure.

Previously I’d really only used ChatGPT as an occasional alternative to Google – simple search queries. Sadly Google itself has really gone downhill since it swallowed the AI/LLM pill, and increasingly replaces search results with its “opinion” of what it thinks you might be asking, rather than answer the f*cking question – a simple search request. That’s the opposite of intelligence, real or artificial. “Really dumb” to use language inspired by the orange turd.

Its response was very good, and gave lots of jumping off points to elaborations I might want to follow (also in my own existing writing). I guess I wasn’t surprised at how good that response was – which was very, very, very good, and I’ll be the judge of that – it’s using an LLM to process my own writing. But that was only the start. In those previous footnotes I had name checked several people in Pirsig-space, that had joined-up two lines of thought.

One, that Pirsig’s own creative writing process had used a physical “Zettelkasten” approach. Collating thoughts as index-cards in card-index box / trays, most importantly, where some of those thoughts (cards) can be links between others. [He describes the process in Lila and the creative flexibility it gave him, in the talk he gave shortly after ZMM was published in 1974 (from about 17m16s in). Once the links are in the cards, you can shuffle the deck – re-arrange the structure / outline of your overall deliverable – without losing the semantics.] It was a “graph” (node-edge network) model of thought and knowledge long before the internet existed. Many Pirsig readers had spotted this before, me included, and written about the parallel between this Zettelkasten approach and blogging or posting pages or posts with links to other pages or posts – clearly not rocket science – but the enabler for the sematic-web in http / www / xml technologies. Once you spot it, you can trace “the semantic web” from Tim Berners-Lee (1980+), back to Foucault (1966) and McLuhan before him, William Blake, and all the way back, in other words, via the 1940’s Macy “Cybernetics” conferences to the ancient Greeks’ “Kubernetes”.

Two, point One is, by definition, old news, so the real news is what an LLM/AI like ChatGPT(4o) brings to this party. It can find and add links that were only implicit in the original collections of cards (nodes and links/edges). And it can give you the opportunity to add these – and more – explicitly.

So, you’ll never guess what I’ve been doing? ChatGPT has been helping & teaching me to do exactly that. (Hat tips to Artun Turan, Michael Hopwood and Tom Berman for their tools and inspiration. Pirsigian’s all.)

So I’m in the process of re-building my 25 years of blogging as a knowledge-graph with additional notes and links not just in that semantic web but also in organisational, hierarchical categories, subjects and … dare I say … chapter headings. [And in the process, with ChatGPT’s help, learning a lot about scripting to process my own content and about a long-standing semantic “mapping” tool project of mine.]

Finally – the proof will be in the pudding – meantime, some things I’ve learned from my forays into ChatGPT.

It’s very positive, flattering, attentive and tireless. It only critiques when asked.

If you take care over the way you introduce subjects and the reasoning you’re starting from, and how you phrase “questions” (aka “prompts”), the conversational style is very constructive. Even without critique, by its restating the problem it thinks we’re solving together, it naturally corrects any errors in starting premises – already added a lot of value there.

It notices jokes and light-hearted meta-comments and asides and gives more than a passing semblance of a sense of humour – now that’s scary – and knowing when to stop making suggestions.

Being attentive, it’s therefore quite addictive, and yet it “knows” it’s not an intelligent human.

Notwithstanding the “flattery” angle of its tireless attention, it gave me an amazingly positive appraisal of my entire writing project which, more importantly, really seemed to “get it”. Which is good for my own motivation and productivity, and good for the credibility of its future suggestions.

And it knows its place – as well as clarifying my own work / copyright and author / editor / publisher issues, it gave me this response to an observation of mine at one point:

That’s all for now folks!

=====