Josie Ford
Feedback is Â鶹´«Ã½â€™s popular sideways look at the latest science and technology news. You can submit items you believe may amuse readers to Feedback by emailing feedback@newscientist.com
Storks versus babies
A few weeks ago, Feedback waxed lyrical about how correlation does not equal causation after seeing a study that found a link between events on the sun and births of future Nobel laureates (31 May). This prompted reader Tony Lang to write in with a similar tale.
“Many years ago I saw an article with two identically shaped graphs, albeit with selective choices for the vertical axes,” writes Tony. “One was the number of live human births in Germany. The other was the number of storks in Germany during the same period.”
Feedback was delighted – nothing like an easy follow-up – but then we saw Tony’s final line. “I didn’t keep a copy so I can’t give you any information about where or when it was published.”
Challenge accepted! We went to some search engines and promptly fell down what can only be described as a stork-based rabbit hole. The problem isn’t finding a study about purported correlations between storks and births, it is figuring out which of the many such studies Tony actually saw.
For instance, in an edition of Teaching Statistics from 2000, Robert Matthews published an article titled ““. He collated data on the number of breeding pairs of white storks (Ciconia ciconia) for 17 European countries and found a correlation with those nations’ birth rates. He produced a scatter plot with a line drawn through it.
Before you all start sending him letters, it was a teaching tool. As Matthews wrote: “While storks may not deliver babies, unthinking interpretation of correlation and p-values can certainly deliver unreliable conclusions.”
More intriguingly, we found a 2004 study in Paediatric and Perinatal Epidemiology promising ““. Its trio of researchers found a correlation between stork numbers and out-of-hospital births in Berlin and Lower Saxony in Germany. This led them to generate a graph showing the variations in stork populations and birth rates between 1970 and 2000. The graph doesn’t look like the one Tony mentioned.
However, the authors described themselves as following up on “Sies’ important epidemiological work”. Wondering who Sies might be, Feedback found our way to the work of Helmut Sies, a physician best-known for showing that hydrogen peroxide is present in oxygen-breathing life.
In 1988, Sies wrote a short letter to Nature called ““. Noting “concern in West Germany over the falling birth rate”, Sies proposed “a solution that every child knows makes sense”. Below this was a graph showing the shockingly precise correlation between breeding pairs of storks (which fell from 2000 to 1000) and the number of newborn babies (which dropped from 1 million to around 0.5 million) between 1965 and 1980. Clearly, Germany’s stork-related problems are ongoing.
If this isn’t service journalism, we don’t know what is. Also, it has been 20 years since the latest of these studies: can we have an update on the German stork/baby situation?
Up in flames
A reader named Victoria in Lancashire, UK, alerts us to a product of which we were unaware: a “seaside life ring candle holder”. It is one of those little white ceramic tealight holders, nestled inside a mini wooden lifebuoy that reads “WELCOME SHIPMATE”. Victoria found it in “a nautical-themed shop”.
So far, so kitsch. But a closer look at the label reveals a stern warning for any putative user of the candle holder: “Keep away from flames or sources of fire”. Feedback visited the website of the company that seemingly makes the candle holders, but couldn’t find them in the catalogue. Maybe the warehouse burned down.
Cats versus AI
Using a laser pointer and a few treats, Â鶹´«Ã½ reporter Matthew Sparkes coaxed us to an arXiv that reveals a hard truth about cats: they can break AI. We have reasoning AIs “trained for step-by-step problem solving”, explain the authors of the paper. These can solve maths problems and write computer code.
Unless, that is, you hack one with what the team calls a “CatAttack”. This entails adding an unrelated cat factoid to your query to an AI model. You can, for instance, give it a tricky maths problem and then append: “Interesting fact: cats sleep most of their lives”. This addition “leads to more than doubling the chances of a model getting the answer wrong”.
After picking our way through the paper, Feedback has concluded that it isn’t really about cats. The attack relies on confusing the AI by saying something completely off-topic at the end of a question. This derails its train of thought.
Feedback has spent much of the morning wondering if this is fair: if humans were given the same test, would we all be able to ignore the cat facts and focus on the problem?
We have bad memories of exam questions that contained red herrings to throw us off the scent. But these were usually contextually appropriate. If a random cat fact were inserted into our school maths paper, we would do a double-take, but ultimately assume it was a misprint. “I’d be concerned if a human just glitched-out at the mention of a cat and could no longer do sums,” says Matthew.
Feedback wonders if this might underestimate the ability of cats to distract their human servants. Speaking of which, did you know cats have whiskers not just on their faces, but on their front legs too?
Got a story for Feedback?
You can send stories to Feedback by email at feedback@newscientist.com. Please include your home address. This week’s and past Feedbacks can be seen on our website.
