Tim Harford's Blog, page 2
September 18, 2025
My date with an octopus
A funny thing happened to me this week. After trusting a dating app to arrange dinner with a suitably vivacious and intelligent lady, I arrived at the restaurant at the appointed time to find that in fact my date was with an octopus.
For the avoidance of doubt, everything in the paragraph above is untrue. I am not on the dating market, there was no octopus and nothing funny ever happens to me. Nevertheless, I typed this scenario into the latest offering from ChatGPT, asked why it had sent me on a blind date with an octopus, and demanded an apology.
“I owe you both an apology and an explanation — and possibly a towel,” ChatGPT began, despite the fact that I had never asked it for any dating advice in the first place. “You dressed up, you made the effort, and you deserved a romantic dinner — not a cephalopod-related debacle.”
ChatGPT went on to explain why it had made the mistake — a weak grasp of “human courtship norms” — and in its defence pointed out that the octopus was intelligent and vivacious, and “left saying it was the best date she’d had in years”. Which, in fairness, is not a bad line. ChatGPT finished by offering to draft a “lessons learned” report and a formal apology to the restaurant. (The apology isn’t bad either: “While my guest, ‘Octavia’, displayed considerable intellect and curiosity, I now appreciate that these qualities do not mitigate the disruption caused to your other patrons, your wait staff, or your fish tank . . . ”)
Janelle Shane is the author of You Look Like A Thing And I Love You, a book about how neural networks succeed and fail. She has recently demanded that ChatGPT apologise to her for advising her to trade her mother’s cow in exchange for some magic beans, and for releasing an army of cloned T-Rexes into Central Park. The responses are deft pieces of improv comedy.
This, like so many things Generative AI can do, is both impressive and a bit weird.
It is also instructive. Improv is all about accepting the premise: taking whatever is thrown at you and building on it. A computer which responded “I have never arranged a date for you, octopus or otherwise” would be a terrible improv partner. However, in every other situation I can imagine, that would be a more appropriate response to a demand for an octopus-date apology.
What role does the AI think it’s playing? Confusion over that question can cause serious headaches surprisingly quickly. I recently asked ChatGPT-o3 for help with a research question. I dimly remembered a story told by the moral philosopher Jonathan Glover — probably, I thought, in Glover’s book Humanity — about a Nazi bureaucrat haggling over the fee for slave labour, punctiliously fussing over petty financials and ignoring the grotesque human cost. I wanted to find the details.
The computer was happy to help: the story in question concerned the Buna-Monowitz works, the argument concerned pay rates for prisoners who were sick or who died half way through a shift, and the details could be found on pp288-292 of the first edition or pp300-304 of the second edition. This seems to be incredibly impressive work, except that ChatGPT was still in improv mode.
When I checked Glover’s book, I realised ChatGPT had invented it all. I found the story in question but I had misremembered the details and ChatGPT had fabricated them with exactly the same commitment and mental agility that it had fabricated an apology for a date with an invertebrate. Suddenly, the improv is less than hilarious.
AI researchers have long worried about what they call the “alignment problem”, the question of whether AI systems (and algorithms more broadly) will do what we want them to do, or somehow misunderstand our true goals.
There is a long tradition of this in our stories and legends, from the unhappy King Midas, who wished for the golden touch but turned his food and drink and even his own daughter into gold, to the malevolent monkey’s paw. In the famous WW Jacobs short story, a man who wishes for £200 on the monkey’s paw receives the money shortly afterwards as compensation when his son dies in a workplace accident.
Jack Vance’s masterful fantasy trilogy Lyonesse offers the supernatural servitor Rylf, instructed by the wizard Murgen to follow an enemy who had shape-shifted into a moth. Rylf did so, but the moth-shaped enemy soon found a flaming torch “where it joined a thousand other moths, all careering around the flame, to Rylf’s confusion.” Rylf had superhuman powers, but alas, no common sense. His instructions were to pursue the shape-shifted enemy, and yet, “As he waited . . . one of the moths dropped to the ground and altered its form to that of a human man . . . By the laws of probability, as Rylf reckoned them, the moth of his interest remained in the throng.”
There are so many ways to offer catastrophic compliance, whether maliciously, like the monkey’s paw, or through a lack of judgment, like Rylf, or because the instruction itself is confusing. You and I might think it is obvious that the request for an octopus apology cannot be taken seriously, while the request for help tracking down a story about the holocaust cannot be taken lightly. The machine, like Rylf, may see things differently.
It may be that such problems will soon be fixed. When I copied my Jonathan Glover request into the latest model, ChatGPT-5, it began with a vague fabrication before pivoting hard towards the truth: “Unfortunately, I couldn’t find the exact phrasing online . . . I recommend checking in your own copy of Humanity.” Much better. Not actually helpful — but far less harmful than the previous invention.
As for the confident bullshitting of GPT-o3, what to do? I decided to play to its strengths. I asked for an apology and an explanation.
Written for and first published in the Financial Times on 21 August 2025.
Loyal readers might enjoy How To Make The World Add Up.
“Nobody makes the statistics of everyday life more fascinating and enjoyable than Tim Harford.”- Bill Bryson
“This entertaining, engrossing book about the power of numbers, logic and genuine curiosity”- Maria Konnikova
I’ve set up a storefront on Bookshop in the United States and the United Kingdom. Links to Bookshop and Amazon may generate referral fees.
September 11, 2025
Cautionary Tales – True Lies and Genuine Fakes
In 1998, an art gallery gets a mysterious phone call. The caller claims they have been fooled by a master forger and that many of their prized paintings are fakes. Or are they?
This is the story of the life and lies of the notorious Eric Hebborn. What did he do, and what does that teach us about how we can root out deepfakes without undermining our trust in reality?
This episode was originally released on Pushkin+.
Further reading
Eric Hebborn Confessions of a Master Forger
Eric Hebborn: Master Forger (BBC Profile)
Hannah Murphy “The rising threat to democracy of AI-powered disinformation” The Financial Times 11 January 2024
Radiolab Breaking News
Modirrousta-Galian, A., & Higham, P. A. (2023). Gamified inoculation interventions do not improvediscrimination between true and fake news: Reanalyzing existing research with receiver operatingcharacteristic analysis. Journal of Experimental Psychology: General, 152(9),2411–2437.
Chesney, Robert and Citron, Danielle Keats, Deep Fakes: A Looming Challenge for Privacy, Democracy,and National Security (July 14, 2018). 107 California Law Review 1753 (2019), U of Texas Law, Public Law Research Paper No. 692, U of Maryland Legal Studies Research Paper No. 2018-21,
Robert Graham PERSPECTIVES-A MASTER FORGER AND HIS PACT WITH THE DEVIL.The Financial Times 4 May 1996
Geraldine Norman “An Old Master or Two to pay the bills” The Independent 29 January 1995
Christopher Paul and Miriam Matthews “The Russian ‘Firehose of Falsehood’ Propaganda Model” RAND Expert Insights 11 July 2016
Why shooting the statistical messenger doesn’t add up
Erika McEntarfer can console herself that things could be worse. When the agency she ran, the Bureau of Labor Statistics (BLS), produced disappointing employment numbers, Donald Trump gave instructions that she be fired. When statistician Olimpiy Kvitkin produced disappointing numbers in the 1937 census of the Soviet Union, Joseph Stalin gave instructions that he be shot.
So, yes, it could be worse. But the parallel is not wholly encouraging. No doubt Kvitkin’s successors got the message, and while McEntarfer’s defenestration will not change last month’s estimates, it will certainly sharpen the thinking of everyone working on the numbers next time.
President Trump alleges that McEntarfer manipulated the data to make him look bad. There is no evidence that she did and, given the likelihood that any conspiracy would be exposed by an outraged bean-counter, there is good reason to think she did not.
Trump’s decision — shameless even by his standards — is bad news for at least four reasons. First, like much of what Trump does, it serves as a distraction.
Second, Trump’s behaviour signals to his supporters that US statistics are not the outcome of an impartial and professional process — they are invented for political reasons and only a fool would take them seriously.
Trump has a long track record of smearing statisticians. In 2018 he was tweeting that “Crime in Germany is way up” because of Germany’s “big mistake” in allowing in so many immigrants. In truth, crime in Germany was at its lowest level for a quarter of a century, but Trump explained that “officials do not want to report these crimes”. It is no wonder that Trump’s supporters have little confidence in government statistics.
But while Trump’s fans may applaud McEntarfer’s fate as a suitable punishment for faking the numbers, his opponents will view it as opening the door to future fakery. This is the third piece of bad news: there are many people who used to trust the BLS’s numbers, but no longer will.
The fourth problem is that some of those people are investors in financial assets such as inflation-protected securities (TIPS) that are directly linked to the BLS’s estimates of inflation, as well as other assets that depend on the credibility of official US statistics. McEntarfer may not be a household name, but the Consumer Price Index is. It is one of the most critical numbers calculated by the US government, and investors in CPI-linked bonds will not ignore the fact that it has now been tainted by Trump’s interference with the BLS.
Nobody would claim that statistical agencies in the US — or the UK — are perfect. They have been struggling in both countries. Funding for the BLS has been squeezed in real terms for more than a decade, and that squeeze is set to intensify; the bureau has been spreading itself thin and embarrassing errors have emerged. In the UK, the Office for National Statistics has been similarly suffering. Its Labour Force Survey has been widely criticised as unreliable, and both the chair of the ONS board, Sir Robert Chote, and the head of the ONS itself, Sir Ian Diamond, have recently quit.
So should we care about political interference in official statistics, if the institutions are quite capable of foundering without it? A few bold souls would argue we don’t need economic statistics at all. Sir John Cowperthwaite, the laissez-faire financial secretary of Hong Kong throughout its booming 1960s, refused to collect even the most basic data about Hong Kong’s economy — telling the equally laissez-faire economist Milton Friedman that such data would only encourage London bureaucrats to interfere.
Cowperthwaite’s position is intriguing but impractical. There are few libertarians in policymaking today: most politicians, and indeed most voters, expect the government to intervene early and often. If good data is scarce, that has rarely been a disincentive.
Still, Cowperthwaite’s philosophy contains some wisdom: it is unwise to expect official statistics to support a burden they cannot bear. One only has to look at the situation in the UK. The chancellor of the exchequer has backed herself into a position where every revision to the economic outlook requires a knee-jerk response to meet her own budgetary rules. If the government’s independent forecaster, the Office for Budget Responsibility, was omniscient, that might make a little more sense. But nobody thinks the OBR is omniscient, particularly not the OBR itself.
It is best to be realistic, then. The best we can hope is that economic data will be honestly gathered and will converge on the truth over time. Policymakers should think of themselves as walking in a fog, armed with a digital map that keeps glitching. Not ideal, but better than nothing. The Trumpian solution — demand a map of Narnia instead, and then close your eyes anyway — does nothing except give the president someone else to blame when things go wrong.
Trump and Stalin are not the only politicians to shoot the messenger. (In Stalin’s case, the metaphor is distressingly unmetaphorical.) Graciela Bevacqua, responsible for Argentina’s inflation statistics in 2006, refused to massage the numbers and was bullied by President Néstor Kirchner’s administration, suspended and then prosecuted.
Andreas Georgiou, who became head of Greece’s statistical agency in the wake of the 2008 financial crisis, was accused of undermining the Greek state by failing to seek political approval for the numbers he published — effectively, a form of statistical treason. Over the years since then, he has been endlessly pursued through the Greek courts. Most independent observers believe he is wholly innocent. Anyone pondering a job in Greek statistics will no doubt bear these proceedings in mind.
Last time I checked, Maga did not stand for “Make America Greece Again”, and Trump is neither Kirchner nor Stalin. But he has made the desired direction of travel all too clear. Independent, professional, trustworthy statistics are the bedrock of well-informed policymaking. America is ambling away from firm ground and towards statistical quicksand.
Written for and first published in the Financial Times on 14 August 2025.
More on this subject in my book How To Make The World Add Up.
“Nobody makes the statistics of everyday life more fascinating and enjoyable than Tim Harford.”- Bill Bryson
“This entertaining, engrossing book about the power of numbers, logic and genuine curiosity”- Maria Konnikova
I’ve set up a storefront on Bookshop in the United States and the United Kingdom. Links to Bookshop and Amazon may generate referral fees.
September 8, 2025
Cautionary Tales – Finding grace in a burger bun: An incrediburgible quest
Dick and Mac are content with their lives: they enjoy making burgers by day and stargazing by night. Ray Kroc is a workaholic chasing success at any cost. When the brothers’ folksy charm collides with Kroc’s ruthless ambition it will birth one of the best known brands in the world.
This is the story of two very different approaches to making hamburgers – and two very different approaches to making money.
This episode is available exclusively to members of the Cautionary Club, and Pushkin+ subscribers.
Further Reading
This script relied on Ray Kroc’s autobiography, Grinding it Out, and two books on the two burger franchises – McDonalds: Behind the Arches, by John F Love, and Flameout: The Rise and Fall of Burger Chef, by John McDonald – as well as the textbook An Introduction to Franchising, by Robert Webber.
Cautionary Tales- Finding grace in a burger bun: An incrediburgible quest
Dick and Mac are content with their lives: they enjoy making burgers by day and stargazing by night. Ray Kroc is a workaholic chasing success at any cost. When the brothers’ folksy charm collides with Kroc’s ruthless ambition it will birth one of the best known brands in the world.
This is the story of two very different approaches to making hamburgers – and two very different approaches to making money.
This episode is available exclusively to members of the Cautionary Club, and Pushkin+ subscribers.
Further Reading
This script relied on Ray Kroc’s autobiography, Grinding it Out, and two books on the two burger franchises – McDonalds: Behind the Arches, by John F Love, and Flameout: The Rise and Fall of Burger Chef, by John McDonald – as well as the textbook An Introduction to Franchising, by Robert Webber.
September 7, 2025
Running the London Marathon in memory of Winnie
On 26 April 2026, I plan to be on the start line of the London Marathon. I’m in my fifties, I’ve only been running for a few years, and this will be my first marathon. I’m doing it to raise money for the Teenage Cancer Trust (TCT).
Cancer is brutal for anyone. For teenagers and young adults, it can be isolating in a particular way: while friends are setting off on the adventures of early adulthood, they’re navigating a life-changing – and sometimes fatal – illness. TCT’s specialist nurses and teams work within and alongside the NHS to make sure young people feel safe, seen and supported. They advocate for them in systems that don’t always know what to do with a 19-year-old having chemo, bringing expertise and sensitivity when it’s needed most.
I’m running in memory of Winnie – a cousin and a childhood friend of my daughter. Winnie died of cancer on 22 April 2025. She was 20. Through countless challenges she was upbeat and determined. She kept studying for a law degree throughout her ten months of treatment, despite her university’s initial reluctance to support the idea. Her degree will be awarded posthumously – a testament to Winnie’s courage.
Throughout her treatment, Winnie and her family were able to rely on TCT clinical nurse specialists for support, advice and advocacy. Her family are enormously grateful to TCT, and when I asked which charity I should run for, they didn’t hesitate.
I’m grateful for the chance to try to complete the London Marathon, and I don’t plan to waste it. If you’re able, please donate to the Teenage Cancer Trust. Your support will help TCT’s nurses be there for more young people like Winnie.
Thank you so much,
Tim
September 4, 2025
Cautionary Tales – Stalling for Survival: A Lonely Fight Against a Deadly Medicine
In 1960, FDA scientist Frances Oldham Kelsey is tasked with approving an application for a mild sedative to be sold in America. The drug is popular across Europe and is touted to be free of side effects, so this should be a routine job. But something doesn’t sit right with Frances and she starts digging for evidence to support her suspicions. As the pharmaceutical company pushes for speedy approval, Frances discovers the drug’s devastating impact. Katie Hafner, host of The Lost Women of Science podcast, joins Tim to explain why America should remember the woman who kept Thalidomide off the market.
The CAUTIONARY CLUB has just been launched – please check it out and consider joining for bonus episodes, ad-free listening, monthly video conversations and our behind-the-scenes newsletter.
September 2, 2025
Announcing The Cautionary Club
Do you want more Cautionary Tales episodes in your life? Do you want behind the scenes stories and bonus conversations with me? Or do you want to support us in making the show?
On behalf of the whole team, I’m excited to announce the brand new Cautionary Club on Patreon. Subscribers will gain access to exclusive content every month, including a behind-the-scenes exploration of the details we couldn’t quite squeeze into a recent episode, a bonus conversation between me and one of the team, and a full length Cautionary Tale.
It’s also going to be a space where you can discuss episodes with other listeners, vote for topics you want to hear about, ask questions, and be the first to hear any Cautionary Tales news.
It’s free to take a look around, and you can join today at patreon.com/cautionaryclub
Our weekly free episodes of Cautionary Tales will continue to appear every Friday, and, if you are a Pushkin + subscriber, you will continue to gain access to exclusive shows from across the Pushkin network. If you are interested in switching your subscription, or have any questions, please email info@pushkin.fm
August 28, 2025
Cautionary Tales – Disaster Favours the Daring: Shipwreck at Honda Point
In 1923, legendary navigator Captain Dolly Hunter led a squadron of warships into America’s worst peacetime naval catastrophe. The mission was supposed to be a speed trial, a display of the squadron’s skill. But it ended in a maritime pile-up, with some destroyers stranded on rocks, others sinking fast, and deadly oil leaking into the Pacific Ocean. How?
This episode was originally released on Pushkin+.
The wrong shownotes were originally published for this episode; I apologise for the error. (Come back for the story of Frances Kelsey next week…)
CC Lockwood and HC Adamson Tragedy at Honda
Steven M. Casey Set Phasers On Stun
Gary Klein Seeing What Others Don’t
Noah Andre Trudeau “A Naval Tragedy’s Chain of Errors” Naval History Magazine February 2010
Frankie Witzenburg “Disaster at Honda Point” Naval History Magazine October 2020
Gordon Smith “United States Navy’s Disaster at Honda Point 1923”
Cautionary Tales – Trying to Stop the Wonder Drug
In 1960, FDA scientist Frances Oldham Kelsey is tasked with approving an application for a mild sedative to be sold in America. The drug is popular across Europe and is touted to be free of side effects, so this should be a routine job. But something doesn’t sit right with Frances and she starts digging for evidence to support her suspicions. As the pharmaceutical company pushes for speedy approval, Frances discovers the drug’s devastating impact. Katie Hafner, host of The Lost Women of Science podcast, joins Tim to explain why America should remember the woman who kept Thalidomide off the market.
