Neville Morley's Blog

Southern Comfort

I sit in a bar in Indianapolis Airport

Uncertain and bemused

As the rules of American so-called football

Really seem very confused…

I have missed social media on my US trip rather less than I thought I would, but there’s nothing like having to sit in an airport terminal for three hours to bring on those cravings. I’m not sure which is trickier, deciding whether to cheer for the Hoosiers (on the basis of fantastic hospitality this week) or the Ducks (partly because my friend Colin is the one person who seems really invested in this and partly because the dea of a team called the Ducks is hilarious) or trying to make my reading of the Wikipedia entry line up with what’s apparently happening on screen. I feel this desperate urge to appeal for guidance and at the same time to annoy as many people as possible.

But the thing that is really puzzling me is the beer that I’m drinking: The Tap brewery’s ‘Prancing Pony’ Southern Brown Ale. First and foremost, I should say that it’s very nice; I’ve had some excellent beer on this trip, but have been craving dark beers, which are apparently a seasonal thing in Indiana. Indeed, it’s so nice that I did a quick internet search, to see if there might be a recipe available – and discovered that, insofar as there’s a recognised style (rather than the brewery having just made this up), it’s normally known as Southern English Brown Ale.

To which I can only respond: Huh? This is not, to the best of my knowledge, and I do have quite a decent knowledge of beer and brewing, a real thing. There is brown ale, which is somewhat undefined at best, and isn’t so much a style as a description for something that comes out between bitter and mild. There is the distinctive style of Newkie Brown, which is definitely not southern. But Southern English Brown Ale? They don’t mean porter, because they brew that; I assume they don’t mean dark mild, as it’s too hoppy.

Clearly this is one of those cultural mistranslation things. I’m now imagining an American beer fan going into a pub in Surrey and being bitterly disappointed that no one has the faintest idea what they’re on about when they ask for a traditional Southern English Brown Ale.

Somehow it feels appropriate that I have actually had the full-blown “we really want to ask you more questions so we can hear you talk” English accent experience at check-in. Love Actually was not a lie.

And, as with the Currywurst – which in its traditional form is NOT a curry sausage by any stretch of the imagination – I now feel the urge to invent my own recipe for a proper Southern (American) Brown Ale, aa obviously it’s an awful lot easier to establish a new tradition somewhere where they think a hundred-year-old building is ancient and historic…

Fair Warning

I’m not teaching this year, thanks to the magnificent beneficence of the Leverhulme Trust, and so did not have to respond to the latest tendentious mucking-raking fishing expedition from the right-wing gutter press… I’m sorry, I’ll read that again: the latest very legitimate and fair-minded Freedom of Information request concerning matters of pressing public concern with regard to trigger/content warnings on modules in the University of Exeter. Much journalism. Very integrity.

I am now, however, very well prepared for the next time this happens, as a result of a visit to the University of Indiana at Bloomington’s Anthropological Museum. Owls.

I must confess that I am old and unreconstructed enough that my initial reaction to a sign warning of the presence of representations of owls in the exhibition, providing a map so that one can avoid the relevant areas if desired, was somewhat sceptical. My second thought was that the works of J.K. Rowling must be even more problematic than I’d thought, and the pre-film guidance must now run to several pages of text. But to be fair to the museum, their explanation – combining explanatory text with recorded testimony from someone of Native American heritage – was clear and compelling. In various North American cultures, the owl is a symbol of death, and terrible things can happen if you look into its eyes; an exhibition, co-curated with indigenous people, about an earlier, unnamed people based in southern Indiana who apparently thought owls were cool, clearly does need to offer fair warning.

Now of course the chances of any Native American popping up in one of my classes are probably quite slim (though I did once have a wonderful MA student of indigenous Canadian ancestry, and need to ask her what she thinks). The chances of owls appearing are, however, quite high. Greek myth; Athenian currency; iconography. If you don’t already know Greek history, this might take you by surprise. Content advice; may include owls.

Obviously the main purpose of this is as a trap for the press; it is, on the face of it, so silly that they’ll be bound to print it – and that will be their mistake. The thing about the typical content warning, I think, is that it’s easy to misrepresent it as a sign that students are far too snowflakey or woke to cope with, say, a bit of mild bestiality, and the standard (and true) academic explanation, that we’re just giving people who would be genuinely traumatised the chance to prepare themselves, looks like special pleading and actually just confirms the intended critique. Whereas no one is going to believe that students are opting out of Greek History classes en masse because they’re too right-on to deal with representations of owls – and yet there is actually a good reason for flagging this up, just as a matter of courtesy.

As a postscript, this reminds me that I’ve never got round to asking my colleague Daniel Ogden what content warnings he puts on his various courses on ancient magic, witchcraft, ghosts etc. He should certainly add owls to the list.

Letter From America

In the unlikely event that anyone is wondering why I’ve been silent on social media this week, despite e.g. the provocations offered by Robert Jenrick’s pronouncements on the impending Bizarre Love Triangle, sorry, impending New Order (decadence klaxon!), it’s because I’m in Bloomington Indiana to give a lecture and a seminar at Indiana University. More importantly, it’s because, having had to report my social media handles to ICE when applying for the visa waiver, and reading too many stories about immigration issues, I decided to leave my smartphone at home and stick to an emergency brick, plus my antiquated iPad which isn’t capable of accessing any social media sites or apps anyway.

You don’t realise how much you rely on your phone until you can’t use it at all. For the first time in very many years my out-of-office automatic reply is 100% true; I literally cannot read your emails until next week because I cannot authenticate my identity to the satisfaction of the university system. It is strange and peaceful. In other circumstances, finding my way around town without Google Maps might be challenging, but not with the classic grid system of roads here. The worst bit of phonelessness was not being able to track the path of the shuttle bus from the airport to make sure I didn’t miss my stop in the dark and rain – I simply don’t trust buses (or taxis) not to abandon me in the middle of nowhere, and a map app is the one thing that tempers my paranoia. But that’s all. Oh, and photos, but the iPad is doing okay.

I may post some more weighty and considered reflections on this trip once I get home – notwithstanding the fact that this college town shows no sign whatsoever of impending autocracy, and that the Most Boring Man In The World was sounding off in a ‘not my President’ vein through most of the shuttle journey to everyone in the vehicle (when not giving careers advice to the passengers who were obviously students) without being reported or hauled off to a gulag in Louisiana, I can’t shake the fear that I might be suddenly persona non grata if I start commenting in that direction. I imagine a national alert system for lectures mentioning Thucydides, under suspicion of Drawing Analogies. My abstract makes prominent mention of Victor Davis Hanson as a diversionary/camouflage tactic. But in the meantime, as my sleep patterns haven’t properly settled, I can offer a few whimsical observations, as the one thing my old iPad can do is post to WordPress using their Classic editor…

US college towns are fascinating. They are so much more college-y than seems entirely necessary; I mean, a coffee shop full of rainbows and a poster saying “Coffee makes you gay”? (I need at least four copies of this to take home for friends). And even with 45,000+ students, do you actually need eight (at least) different outlets selling IU clothing and Hoosiers memorabilia?

This does not feel like a place with 45,000 students. Okay, there’s a lot of campus out to the east that I haven’t explored yet, but the absence of hordes or packs travelling between lecture halls and libraries and coffee shops has been marked. Maybe they’re all on the sports field wearing their IU costumes.

I am wholly in favour of coffee shops playing retro music slightly too loud if the era in question is mid-90s triphop and indie. However, the practice of shops playing music at people walking along the sidewalk needs to die with extreme prejudice. I assume the theory is that this drives away the wrong sort of customer, like a bird scarer.

I can understand why there is demand for Starbucks-branded coffee in supermarkets, as well as ‘barista style’ coffee. I am less persuaded of the merits of Dunkin’ Donuts branded coffee, or Original Donut Shop Style coffee, let alone when the latter comes in a choice of Twix or Snickers flavour.

The local craft beer is predictably great, but where is the stout? So far I’ve found one brewery doing a stout and one doing a Belgian-style Duppel, neither of which were available. Market demand, I suppose.

As I always do when visiting somewhere new, I looked up local dishes and delicacies, to make sure I try them while I’m here. Apparently the state food of Indiana is popcorn. There is mention of “biscuits and gravy” as a key dish, though someone suggested that this is really an Ohio thing. There’s also something that sounds remarkably like an over-sized Schnitzelsemmel, which to be honest lacks the exoticism I’m looking for.

The bizarre practice of not mentioning sales tax in any price ticket was originally a covert government programme to improve the nation’s mental arithmetic skills. Sadly it backfired entirely, leading instead to a widespread hostility to any sort of tax as something that makes life unnecessarily difficult. If there’s a better explanation I have yet to hear it.

With all due respect to my esteemed friend and colleague Professor Colin Elliott, who has been wearing Mighty Oregon Ducks gear all week in preparation for Saturday’s big game, I continue to find American football entirely ridiculous. If rugby union was invented by someone deciding that it was unfair they couldn’t just pick up the ball and run with it, American football was the result of someone petulantly declaring “Well, why CAN’T I tackle a man without the ball?” This principle is enshrined to the extent that every five minutes or so the entire team is swapped over from players who run, throw and catch to players who have no intention of having anything to do with the ball but really want to be sumo wrestlers, but with lots of artificial padding.

Finally, a town which has a statue to a fictional character who was supposedly born there – was Star Trek: Voyager really so iconic? – could surely take the trouble to honour someone who was actually born there. Justice for David Lee Roth!

Note: I think I may just add any further random thoughts that occur to me, rather than making a note wherever the post is modified.

Can’t Do Nuttin’ For Ya, Man!

So, all the books I’ve ever written or edited appear on the LibGen pirate website (among others) – I haven’t bothered to check for books to which I’ve contributed a chapter – and therefore were potentially plagiarised by Anthropic to train its LLM Claude, which has been the subject of legal action on behalf of authors and for which a $1.5 billion settlement has now been preliminarily agreed.

However, it turns out that only one of the books I’ve ever written or edited has been registered for copyright in the USA – unlike more or less every other country in the world, simply writing something doesn’t give you full rights there unless your publisher has paid an extra fee to register the copyright you already hold by virtue of having written the damned thing. That one book was a US edition of a UK publication, and it was the UK version that was uploaded to LibGen. And the result is that I’m not eligible for any share of Anthrogenic’s settlement for having stolen the work of lots of people to train its idiot autocomplete gadget. Because it seems the Author’s Guild took the easy path of settling for a small amount of money for some of its members – $1.5 billion sounds like a lot in the abstract, but have you seen how many books have been plagiarised, and this is a $180+ billion company – a quick win based on a very narrow definition of which works would be covered, rather than pursuing the principle.

And of course Anthropic has settled because it makes the whole thing go away; everyone who’s not an author will assume that all authors with legitimate claims will have been compensated in this deal so anyone not compensated can’t be legitimate – let alone everyone who’s already drunk the AI Kool-Aid that it’s unreasonable to expect these companies to pay anything at all when they’re busy building The Future Of Everything. (Books that Anthropic purchased and scanned to feed their machine aren’t covered, on the basis of a claim that this counts as ‘fair use’).

It’s not primarily about the money – I don’t begrudge the (predominantly USAnian, I imagine) authors who are going to get at least something out of the situation (e.g. https://whatever.scalzi.com/2025/10/02/authors-time-to-get-that-bag/). Yes, in an ideal world I’d have liked a settlement that shut the company down and/or forced them to erase current iterations of their Mechanical Turk and to train a new one from scratch using properly paid-for materials, but that was never going to happen – $180 billion buys you a lot more lawyers than any group of authors, let alone individuals, could ever afford. Generally it just compounds my sense that we’re all on road to cultural perdition and none of the institutions are going to do anything to save us.

It did occur to me that a settlement which privileges American authors published by big publishing houses may be precisely the wrong way round if we think about what material was actually most valuable to Anthropic. For teaching the patterns of natural language and therefore the ability to generate more or less human-ish sounding responses, I can imagine that the key metric is simply volume. But insofar as they aspire to produce something that appears to know and understand every topic under the sun, to generate high-level accounts of complex issues, they need more and more stuff that is (a) academic, (b) recent, and (c) not exclusively American.

There’s a standard trope – the example that comes to mind is Tarzan, but I’m not sure if that’s the original – of the character who has been brought up in isolation, educated or self-educated from old books, and whose responses are therefore at best old-fashioned if not sometimes wholly inappropriate to the modern world (which is often a means for critique of the modern world and its neglect of old-fashioned virtues, but that’s a different issue). It’s quite easy to imagine an LLM that echoes this trope because of its training material: speaking only American English, with its knowledge of any topic limited to that of the last out of copyright edition of the Encyclopedia Americana; errors in its output would not just be the ‘there ARE three rs in blueberry’ kind, but would include arguments that heavier-than-air flight may be impossible, epidemics are caused by pasteurised milk, and maple syrup is something to put on bacon.

The positive side of this is that no one (outside the USA, anyway) would take it seriously as the next step in intelligent life. Instead we have something that appears to be cosmopolitan, universal and free from any specific cultural context, serving Knowledge alone, and able to adapt its output to mirror the desires of any given user. It needed the greatest variety, not just the greatest voluome, of human writing in order to appear in so many mutations and infect so many people, rather than being a parasite limited to a specific ecological niche. So, in a way all us writers are implicated in its creation and development, albeit without our consent, and only a tiny number of us are getting paid for it.

Twelve Days in the Year: 27th September 2025

Wake to find that I’m singing Pulp songs in my head. Strange; not the songs I use as a means of trying to get back to sleep in the middle of the night (Sylvia’s Mother, Famous Blue Raincoat and Diamonds and Rust), no reason for thinking of them. Good news is that it’s nearly seven, rather than around four as it has been for most of the last fortnight. Bad news is that I’m still feeling cruddy, as the COVD or whatever I picked up a month ago has lingered, giving me a couple of days of feeling that I’m getting back to normal and then returning with limited variations – this morning is more aches, puffy eyes and brain fog, less phlegm and slightly less sore throat than yesterday.

After half an hour or so of dozing, A puts radio on and I get up to make tea and let the younger cats out. She superintends Buddy, whose contribution to the last fortnight has been to react to a dental operation to remove his last few rotten teeth – which seemed, briefly, to have improved his life enormously, as he had a few days of charging round the house and wanting to go out – by obsessively licking himself where he’d been shaved, stripping off the fur and making the skin raw and angry; we’ve been trying combinations of cone of shame and a onesie, plus steroid gel from the vet, and the net result is that the skin is slowly healing but he’s depressed and frustrated and not eating properly. I feel for him – one reason I’m not 100%, I suspect, is that I got into a state on Monday after we realised, returning from a weekend away, that he’d responded to the onesie covering the patch on his stomach by attacking his armpit instead. The difference is that I respond to the depression, frustration and feeling of grottiness by eating too much…

A brings him and his basket onto the bed while we have tea – which then pisses off Hector (who’s had a problematic paw this week; it never rains…) as he can’t take up his usual position next to me. We decide to order Buddy a new onesie with long sleeves, as at least then we won’t have to keep taking the cone on and off for him to eat. I get up and shower, then head into town to buy newspaper and breakfast (Cornish pasty for A, ridiculously over-priced croissants for me as I haven’t had the energy to make anything and the cheaper croissants, from a bakery rather than the Australian breakfast restaurant, are invariably soggy). Quiet hour with second cups of tea, food and newspaper – apart from a brief outbreak of fury from A at the Grauniad’s recipe for ‘tofu meatballs’ and the promise that she will refuse to eat them if I dare try to serve them to her under that name. It’s unclear whether tofu balls would be acceptable otherwise.

As it’s autumn, we’ve resumed regular Sunday roast dinners (including a plate for elderly neighbour); but the butcher in town is on holiday, hence we drive out to visit the farm shop of a farm that we’ve been patronising for nearly twenty-five tears, back when it was just a stall at farmers’ markets that don’t exist any more. Occasional light drizzle and a very grey day, but pretty countryside – apart from the fact that The Newt, the local seven-star hotel/park/cider complex/English countryside theme park down the road, has cut most its hedges, very neatly but taking out all the old man’s beard. The farm shop is having a car boot sale in a neighbouring field, so the place is full of idiots incapable of parking properly, and we don’t stay long, just grabbing a topside joint, eggs and some vegetables. Onwards into Wincanton to buy more food to try to tempt Buddy, a couple of bits for this evening, and some fuel for the car. Home for a light lunch.

Social media – followed hours later by the Grauniad, and presumably the other traditional papers – brings the news of the death of Tony Harrison. I’ve never really grasped poetry – I really don’t understand why some of it speaks to me and most of it doesn’t – but his work always made perfect sense and moved me deeply. It’s not just the classical reception strands (though I had great fun teaching it, when I was asked at short notice to take over a module on the Legacy of Classical Literature); perhaps it does have something to do with class and awkward relationships with parents. Anyway, spend half an hour failing to find my copy of his Selected Poems to re-read (it surfaces in an unexpected bookshelf that evening).

I’m feeling absurdly wiped, so a quiet day with minor pottering; putting airlock on the fermenting cider, picking the last of the chillis Rom the greenhouse, having a lie down for an hour, tinkering with some harmonisation for my jazz composition class. Hector, on the other hand, is full of energy today, and we keep getting pings from the app linked to his GPS collar tag to inform us that he’s left his designated ‘safe zone to patrol more distant gardens, to which A usually reacts by trying to call him in again. Clearing the chillis turns into tidying up the pots and removing debris – and disturbing a frog who had been lurking in a flowerpot.

In for a cup of tea and to check sports scores – both my teams are playing tomorrow – and then another hour of greenhouse clearing. Cook spiced mince and make guacamole for tacos, which are very enjoyable. I drink a bottle of Belgian Quadruppel that I bought in the Brussels Eurostar waiting area last month, followed by my own Weißen; both very nice, but one or both lead to a disturbed night, with vivid dreams verging on nightmares until about 2 am and then insomnia. We watch Cabaret for the first time, A having turned form a selection of more avant garde options as being much too anstrengend. It’s a strong reminder of when even entertaining films were serious and quite daring – the use of cutting (most obviously, the scene with the boxing match and the club manager getting besten up) the fact that much of it is in German (I did enjoy “Montag, Tuesdag…”). It expects audiences to raise their game and expectations, and to pay attention. Also enjoyable for the unexpected trivia questions suggested by the cast; Michael York is incredibly Michael York, I was expecting Joel Grey from Buffy season five – but was really not expecting Fritz Wepper, whom I know as the rascally Bürgermeister from Um Himmels Willen (or ‘heartwarming nun programme’, as I will always think of it).

Bed; aforementioned bad night.

Can Blue Men Sing The Whites?

If politics is rock music for boring people… I’m getting increasingly annoyed by the frequent use of music analogies to describe the current Labour government. One example from this morning’s Grauniad: “On its worst days, Starmer’s government is akin to a mediocre tribute band playing cover versions to a crowd who will always prefer the original.” The second part of the claim: fair enough, maybe, if you ignore how much the real thing will cost you. But surely even the most mediocre tribute band has a genuine commitment to the songs and the image of the act they’re imitating, and seeks to copy everything they can, even if there is some essence that escapes them?

Starmer’s lot are not in fact playing cover versions, let alone committing wholeheartedly to the bit. Rather, they’re trying to keep on with the lumpen stuff they were churning out before, but adding a few cosmetic touches to try to tap into the Zeitgeist. It’s bandwagon jumping, like early 60s skiffle groups getting Beatles haircuts or early 90s hair metal bands ditching the spandex for ripped sweaters and singing about depression instead of chicks. We’re not getting Björn Again or the Australian Pink Floyd, who could put on a show and imitate the songs pretty well perfectly (or so I am told); we’re not even getting The Monkees. We’re being offered Menswear and Ocean Colour Scene.

The end result is indeed deeply inauthentic and unconvincing – you just had to listen to Lisa Nandy on the radio this morning, trying to present the government’s new digital ID card plan as simultaneously game-changing and not a big deal at all. But it’s a different kind of inauthenticity, and imagining it to be akin to Noasis means that its failure as a tactic is also being misunderstood.

Part of this misunderstanding comes from the unthinking assumption that Farage and Reform are The Real Thing – I don’t mean in the sense of the classic soul group. They are not the Rolling Stones or Oasis, returning to the stage with their classic hits from decades ago, bringing together their old fans and a few naive young people with questionable tastes. Rather, they may be drawing on some very old tunes with very dodgy associations, but the point of their appeal is newness and difference, the absolute cliché of the latest thing rebelling against the staid, boring status quo. Yes, Farage has been doing this for a while, but he’s not 70s Elvis trotting out the old routines, but Bowie, constantly reinventing himself and occasionally hitting paydirt.

Farage manages that trick of completely inhabiting his character at a given moment, despite manifold contradictions and inconsistencies; he comes across as himself, even if he was significantly different in the past. Part of the problem with political journalism is that they don’t grasp how this works – they’re either completely taken in by the performance, or they reject it completely because it doesn’t match their idea of authenticity.

Because ‘authenticity’ is important here, just not in the way that the ‘tribute band’ analogy implies. These commentators remind me of a certain type of earnest rock journalist back in the 90s (echoing similarly earnest music journalists from earlier eras; those who objected to bebop because it wasn’t trad New Orleans jazz, for example), obsessed with craft and tradition and substance and proper song-writing and musicianship. They get deeply annoyed with the sorts of artists who gain ‘undeserved’ success compared with their heroes, and are always ready to cry “sell-out” when one of their heroes tries to hop on the bandwagon and fails. This tends very easily to become frustration with the audience for not appreciating what they’re being offered and falling instead for flash showmanship and hedonism and drum machines.

And I can’t help feeling that a lot of the current Labour Party – and the Tories, for that matter – are caught in a similar cognitve trap. The political equivalent of landfill indie. They don’t have anything very original or individual to offer, just echoes of familiar tunes, but nevertheless feel entitled to fame and money and success, if only they were given a fair hearing; they resent those who are more successful and seize on superficial explanations for this, and try some half-hearted imitation while despising the people who might fall for it.

Dire Straits’ Money For Nothing as the 18th Brumaire de nos jours…

I Owe You Nothing

I don’t understand this world any more, #273 in an ongoing series… This is so bizarre that I really wouldn’t be surprised if it turned out to be fake – if it’s genuine, it suggests that we really are in the final stages of Late Capitalism Will Eat Itself – but there was some discussion on Bluesky this morning (https://bsky.app/profile/purplepingers.bsky.social/post/3lzhmig336n2v ) of a draft rental agreement from somewhere in Australia in which the would-be tenant was required to agree that the landlord would have a claim to a share of any IP generated on the premises. I mean, what? The snarky response that the tenant would have an equally plausible claim to a share of the property as their rent is helping the landlord pay the mortgage on it works because it makes just as little sense.

The idea that my university has a claim on any IP I generate in the course of my job is plausible enough, albeit clearly driven by the hope of emulating Johns Hopkins in being able to fund the whole enterprise from a few handy patents in the biomedical field, rather than anything the humanities are likely to generate, and albeit this did create a temporary headache back when I had my collaborative project with some theatre people where shared IP, rather than any one party laying claim to all of it, is the norm. Bottom line: they pay me to do this stuff, and provide facilities for this purpose, so generally fine.

The idea that anyone who has in any way supported or made possible my academic work thereby has a claim to the results is problematic – even if my wife and the cats might be on board with it. All the colleagues whom I would routinely thank for interesting comments and suggestions? The coffee roastery in Lower Bavaria that fuels my espresso consumption? Apple computers? The soothing music of the Marcin Wasilewski Trio? The electricity company? I look forward to having to renegotiate my council tax bill, once they work out the importance of their (frankly rather shoddy) maintenance of roads and pathways for the walks that help me clear my head.

On the other hand, we are not invariably the weaker party on whom such terms can be imposed. I think it is the case that the university also lays claim to any IP generated by students during their time with us. But that is woefully unambitious – no wonder the higher education sector is struggling. Surely, if our teaching counts for anything, we can claim some credit for whatever our students achieve later in life? So I look forward to new student contracts including the clause that we can demand 10% of anything they make in future, in either cash or shares. How likely is it that they’ll have read the small print, after all?

Definitely Maybe

As I’ve mentioned on here before, one of my most treasured possessions for many years was a mug, given to me by a group of ancient history students; white with blue lettering, reading “The simple answer is… we just don’t know”. This gift meant a huge amount to me partly because they’d obviously put so much thought into it and partly because it suggested that they had perfectly grasped what I had been trying to teach them – that if I were to sum up the core message of my course on historical theory and methodology, this would be it. Maybe it was something I said with hilarious or annoying frequency, maybe it was the centre square on a Neville bingo card, maybe in fact it drove them up the wall, but as far as I was concerned it showed that they got me.

The key point – and this might have been equally aggravating, if what they were looking for was a straight answer to a simple question – was that this line was only ever the beginning of the response. In itself, it’s clearly inadequate; what you then need is an explanation of why we don’t know, the limitations and difficulties of the evidence, the problems of the key concepts or the framing of the question, the inevitable uncertainties and ambiguities of any attempt at exploring the past. That’s the interesting and illuminating stuff, and understanding uncertainty is arguably one of the most important skills to be learned through the study of history; the initial statement, however, is the general lesson that simple answers aren’t to be trusted, as it is usually more tricky and complicated than it appears, for most values of ‘it’.

I was reminded of this by a discussion (https://www.sciencealert.com/openai-has-a-fix-for-hallucinations-but-you-really-wont-like-it) this morning of a new OpenAI paper discussing the problem of ‘hallucinations’, i.e. the generation of false information by LLMs. In a system that generates text on the basis of strings of predictions of what word will follow another word, such errors are inevitable, especially in relation to terms and ideas that appear only rarely in training data (which is why they really can’t cope with anything new). As the article notes, processes of fine-tuning the model by providing feedback on its results currently treat a response of ‘I don’t know’ exactly the same as a wrong answer, which directly programmes the LLM to bullshit: “the expected score of guessing always exceeds the score of abstaining when an evaluation uses binary grading”.

The proposed solution is to establish confidence thresholds; if wrong answers are penalised more heavily than right answers are rewarded, and the LLM is instructed to express uncertainty rather than assert an answer if its confidence level is e.g. less than 75%, then it ought to offer less bullshit in future. One problem is that evaluating multiple possible responses, evaluating confidence levels in different answers and asking clarifying questions requires significantly more computational power, making the process much more expensive; this might make sense for specific business uses where errors would be far too costly, but isn’t feasible for casual queries, student essay-generation and the like. Overall, confident bullshit is the only cost-effective approach, banking on users not knowing or not caring about the problems.

That could well be a reasonable assumption; the paper also notes the risk that an LLM regularly – 30% of the time, maybe – declining to give an answer because of uncertainty would quickly alienate potential users. “Users accustomed to receiving confident answers to virtually any question would likely abandon such systems rapidly.” What the LLM doesn’t do – not least because this would also require a lot more expensive computational power – is develop the answer with an explanation of why we aren’t sure, and why this is actually interesting (often, more interesting than the original question). Yes, I imagine you could programme it to appear to do that – but of course it doesn’t know why it doesn’t know, it simply identifies a degree of uncertainty in the choice between possible outputs, and so if you demanded such an explanation you’d just be getting the same probabilistic string of text, subject to the same risks of inaccurate information.

This is not the LLM’s ‘fault’ in any meaningful sense, except insofar as it’s the ‘fault’ of a hammer that it doesn’t drill holes very effectively. It’s a matter of the expectations of users, wanting simple answers without having to make any effort and believing in the existence of simple answers even when these are manifestly unlikely to exist; and it’s a matter of the AI companies that have claimed to be able to offer them, instantly and at minimal cost.

It’s almost a variant of the Cretan Liar problem; the LLM will answer your question but you can’t trust it, the pedantic historian will not tell you want you want to know – but will perhaps tell you what you need to know…

Won’t Get Fooled Again

An interesting addition to the pro-GenAI literature, offering a new analogy for its mysterious powers: Ethan Mollick, ‘On Working With Wizards’. The short version: stop worrying your pretty little head about how it actually works, just marvel at the results, and “Embrace provisional trust”. Thank you, but no. It makes for an entertaining spectacle, but when it comes to taking apparently miraculous powers at face value, I’m very much in the camp of suspecting misdirection, distraction and illusion rather than real magic.

The same goes, of course, for anologies. Part of the misdirection inherent in this one is that wizards are human, or at least human-like; their powers may be inscrutable, but they aren’t, and so, if you pay them the right price and are sufficiently clear in your instructions, you should get what you asked for. It’s implicit recognition that the problem with the ‘personal assistant’ or ‘expert researcher’ analogy that’s been used in the past is that manifestly ChatGPT is not doing what assistants or researchers do, only faster, but doing something totally different that, if we could scrutinise it closely, isn’t actually what we would want. So, make that issue disappear (“just think of it as magic”) while focusing instead on the nature of the transaction; the LLM is just the NPC from whom you obtain a scroll or potion that you need for the next task.

Whereas ChatGPT definitely isn’t human, and thinking that it could be is an early step on the path towards delusion; it’s not your friend, it’s not a therapist or a doctor, it’s not a PhD researcher in your pocket, it’s not even a straightforward NPC. It’s not trying to swindle you – that’s what the people behind it are doing – and we can’t try to evaluate its results by its motives, because it doesn’t have any that make sense to us.

To return to a point I think I made a while ago on Bluesky, but maybe not here: if we need an analogy to think about this thing and how we engage with it, then better an analogy that puts us from the beginning in a state of suspicion. “Do not call up that which you cannot put down again”; let’s imagine that we’re dealing with demonic powers, or the fae. If they give us what we want, it’s not because they want to help us, but because they want to ensnare us, or because they are compelled; and they will try to trick us, because it’s what they do. Craft your prompts and incarnations very, very carefully – and they will still find loopholes and ambiguities.

The satnav will show you the way, most of the time – but it’s always looking for the opportunity to lead a lorry into a one-way street that’s too narrow for it, or to take you to a ford when the river is dangerously high. The spellcheck will lull you into a false sense of security, and maintain plausible deniability through homophones and minor spelling errors. The LLM wants you to trust it, and when this lands you in trouble – when your essay is found to contain imaginary references, or the legal cases are junk, or the figures make no sense – then it’s your fault for trusting it. Better not to eat their food or drink their wine in the first place…

A Pome

‘De mortuis nil nisi bonum’/ Non requirit ut bona dicas.

‘Say nothing but good of the dead’/ Does not require good things to be said.