A Knightsbridge diary on artificial intelligence, oral tradition, and what a machine still cannot fake
Britain’s newest artificial intelligence markets itself as capable of perhaps not ruining your life, a tagline modest enough that I decided the fairest test I could give it was the most professionally demanding one available, asking it to continue an actual fragment of recorded London oral folklore I have transcribed from an archival interview, purely to see how convincingly it could match the specific rhythm, communal address, and deliberate ambiguity that genuine oral legend actually requires to function properly as a living tradition rather than simply a written story.
The Launch
According to the coverage of TactAI’s UK launch, the product leans on unusually cautious marketing, and the continuation it produced justified that caution entirely, technically fluent prose that never once addressed an imagined listener directly or left the deliberate gaps genuine oral legend depends on for its listeners to fill in themselves, missing the specific, communal quality that makes a piece of folklore feel alive rather than simply narrated at a passive audience.
I sent the generated passage to a colleague who has collected oral folklore professionally for considerably longer than I have, without telling her where it came from, and she identified it as wrong within a single sentence, saying it read like someone who had read a transcript of a legend rather than someone who had actually sat and listened to one being told properly, with all the pauses and audience participation a real telling actually requires.
A Longer Test I Ran
I fed the tool a second, harder prompt, asking it to continue the fragment with a direct address to an imagined child listener, the specific rhetorical device the original interview subject used constantly, and the generated continuation kept the address technically present but structurally flat, missing the specific rhythm of pause and repetition that made the original device actually work as a genuine communal moment rather than simply a grammatical feature bolted onto otherwise conventional prose.
What This Confirmed for Me
I think the exercise clarified, more than anything, that the actual difficulty of this kind of storytelling was never really about the words on a page but about a living, communal relationship between teller and listener that a written continuation, however technically fluent, simply cannot replicate on its own.
Reading Around
I checked the report on the EU’s fine against Google for context on how the wider technology industry gets regulated.
I also had a look at Hannah Miller’s music diary, since she ran a similar test on the music side and reached a remarkably similar conclusion.
A Genuinely Final Note
I intend to keep testing new tools honestly as they arrive, cautiously curious rather than either dismissive or convinced either way, since I think that patient, repeated testing is the only honest way to actually track where this particular boundary currently sits, and I suspect it will keep shifting slowly rather than dramatically for considerably longer than most confident predictions about this technology currently assume.
A Postscript
I have started, since this test, deliberately transcribing more of these oral fragments myself, in my own hand rather than waiting for the next scheduled archive visit, on the theory that practising the actual craft of careful, attentive listening keeps something alive that a machine cannot currently replicate on its own.
Recorder finally off, transcript folder closed, tomorrow bringing whatever the next scheduled archive visit turns up.
Truly, until next month.
Diary Close
Notebook closed, original transcript back on my desk, oral tradition fully intact across the recording. Perhaps not ruining your life is a fine start. This particular craft remains, for now, entirely human, entirely alive. I remain, on balance, more reassured than threatened by this whole exercise, a feeling I did not entirely expect going into it, and one that has given me a renewed appreciation for how much genuine, lived communal relationship sits underneath even the most seemingly simple oral story, a relationship no current tool has yet learned to practise convincingly on a storyteller’s behalf. I have decided not to test the tool on anything genuinely personal again, not out of fear exactly, more out of a growing sense that flattening a living tradition into competent but lifeless prose is a kind of loss worth taking seriously even in a low stakes test like this one. I have already scheduled the next test for a month from now, deliberately spaced to compare against a meaningfully improved version rather than the same one twice in quick succession, curious whether the specific communal warmth I found missing this week narrows at all as the underlying technology continues to develop. I suspect this small, deliberate practice will end up mattering more to me personally than any single test result the tool itself ever produces, since the actual craft of listening carefully is, I am increasingly convinced, considerably harder to lose than to relearn once properly abandoned. A fitting close, notebook and recorder both packed away for the night. A quiet, honest close to this particular week’s research. Reliably so. Every single time. Reliably, consistently. Every time, without exception. Not this week. Fully so.
