Cartoonist's AI Experiment Reveals Technology's Hilarious Failures
When satirical cartoonist Martin Rowson decided to test artificial intelligence with a simple personal question, he inadvertently created what he calls the "How bloody stupid is AI?" game. The results were both comical and concerning, exposing fundamental flaws in technology that billions of people increasingly rely upon for information.
The Absurd List of Alleged Wives
Rowson's experiment began when he asked Google's AI overview a straightforward question: "Who is Martin Rowson's wife?" What followed was a parade of completely incorrect answers that grew increasingly bizarre with each query.
The initial response claimed Rowson was married to celebrated lesbian author Jeanette Winterson, an assertion Rowson humorously refuted by swearing "on the lives of the entire population of Silicon Valley" that this was categorically untrue. As he repeated the question, the AI generated an ever-expanding list of fictional spouses.
Among the alleged wives were:
- Textile designer Fiona Scott-Wilson
- Poet Bridget Rose
- Actor Fiona Marr from Bridgerton
- Economist Ann Pettifor
- Various Julia Mills including a fantasy author, illustrator, and late powerlifter
- Writer and journalist Emily Rees
- Lawyer and academic Siva Thambisetty (who AI claimed was married to chess grandmaster Jonathan Rowson, whom it also incorrectly identified as Martin's brother)
- Writer and journalist Carrie McLaren
- Channel 4 News presenter Cathy Newman
- CNN correspondent Clarissa Ward
- Journalist and broadcaster Rachel Johnson (Boris Johnson's sister)
- Rowson's own daughter
Descending Into Complete Fabrication
The situation grew even stranger as the AI began inventing entirely fictional people and relationships. It claimed Rowson was married to "journalist and author Kate Clements Rowson" - a person who doesn't appear to exist in any verifiable records.
Another fabrication suggested marriage to "writer/illustrator Helen Grant" with a son named Leo who was supposedly a jazz musician. The AI even invented a "former Guardian political editor & current CEO of the Joseph Rowntree Foundation Liz Kerr," despite no such person ever holding either position.
Perhaps most bizarrely, the AI created an entire fictional family life, claiming Rowson was married to "Debora Rowson (nee Ffrench)," a retired civil servant, with an additional daughter named Clementine who was supposedly another writer/journalist. It even fabricated a Guardian column where Rowson wrote about their "amusing domestic upsets."
The Real Wife Remains Unknown to AI
Throughout this entire exercise, the AI consistently failed to identify Rowson's actual wife, whom he married in 1987 and who has deliberately maintained no online presence. Rowson notes this is "to everyone's relief," though he finds the situation "very funny" as a satirist who appreciates exposing folly.
Eventually, the AI began responding with "Her name is not publicly named in the provided search results," suggesting either a capacity for learning or simply another form of evasion when confronted with its own limitations.
Serious Concerns Behind the Comedy
While Rowson finds humor in the AI's suggestion that he might be married to Boris Johnson's sister ("imagine that family Christmas!"), he raises serious concerns about what this reveals about AI technology.
"For this nonsense to be the fruit of a garlanded, universal research tool used by billions - and for it to be repeatedly, serially wrong - is more than slightly disturbing," Rowson observes. He notes that AI appears to be "about as sentient as an abacus" and primarily reflects humanity's capacity for self-deception.
The cartoonist warns that combining AI's tendency to "tell humans what it 'thinks' they want to hear" with "the world's most dangerous people" - those who are "idiots who think they are really, really clever" - creates a potentially dangerous combination for society.
Rowson's experiment serves as both entertainment and cautionary tale, demonstrating that while AI might eventually answer "banana bread" or "Exterminate them all!" to complex questions, its current capacity for basic factual accuracy remains deeply flawed and unreliable.
