I kicked off the new year with a fun chore: checking the English translation of my book. The Number Bias will be published this summer, and I’m now ploughing through the text.

It makes me nostalgic. I’m back to a time when I was endlessly digging through thick biographies for that one good detail about Florence Nightingale or that one shocking quotation from Alfred Kinsey. At the same time, it reminds me a lot of the subject I’m writing about now: artificial intelligence (AI).

If there’s a place where AI has made rapid progress, it’s definitely in the field of translation. It’s becoming increasingly difficult to ridicule Google Translate and the like for their wonky translations, even though "Google Translate songs" will probably remain among the YouTube staples for some time to come.

Jimmy Fallon and Idris Elba sing translated (and then re-translated) songs.

"YouTube staples" is literally what I wrote in the Dutch version of this newsletter. Because how do you say "staples" in Dutch? When I translate the phrase in DeepL – a very nice translation programme – I get "nietjes". You know, these bad boys:

Google Translate does it a lot better with "one of the most important things on YouTube".

Still, I wouldn’t write that sentence down in Dutch so quickly. Because, in my opinion, the English "staples" gives a different, more playful feeling. And that is exactly the crux of translation. The meaning can be right, but stylistically it might not fit at all.

It’s not for nothing that I’m working with a professional translator for my book. And those stylistic things are precisely the nicest puzzles to solve.

A lot of hope

Another nice job after the Christmas holidays was reading the contributions under on hopeful developments in AI.

Jaap, for example, was enthusiastic about Transkribus, a programme that digitises old texts, and asked on De Correspondent: "Does similar software exist for handwritten music?" Good question. If you know something, get in touch.

"Loneliness expert" Jeannette thought it was an interesting piece, but she did challenge PhD student Daniela Gawehns’s remark "self-driving-cars … who needs those anyway?" Jeanette writes: "I impatiently look forward to the self-driving car. It can ensure that the elderly and people with mobility problems can (continue to) function independently."

(That reminded me of an article I recently read on Wired: Not because they won’t ever benefit but because the elderly are not involved in the design.)

Machine learning academic Dan responded on The Correspondent with an important critical note: "No matter what good or bad application we have in mind, we should also be aware that intensifying the data collected about vulnerable communities, or rainforests, or clouds leads to a situation where those who control the data have an increasingly strong influence on how our world evolves."


Colleague Gwen Martèl sent me a wonderful clip from The Big Bang Theory just before Christmas. "What’s the best number?" Sheldon Cooper asks. He himself gives the best answer, which, according to himself, is 73.

And I have to agree, 73 is pretty cool. His reasons:

  • It’s the 21st prime number (= a natural number larger than one that you can only divide by itself and by one).
  • Its mirror is 37, the 12th prime number.
  •  21 is the product of – "hang on to your hats" – seven and three.
  • And in the binary system, 73 is a palindrome: 1001001. In other words, if you turn it over, you have exactly the same (just like "a man, a plan, a canal: Panama").
What is the best number? The answer according to Sheldon Cooper.

Thank you, Lètram Newg!

Before you go ...

... of AI reporting shows: "The researchers who appear most regularly in the news are not necessarily the most widely cited by their academic peers and many of them have strong ties to industry."

With thanks to Maurits Martijn, who shared the research on AI reporting.

Prefer to receive this newsletter in your inbox? Sign up here Follow my weekly newsletter to receive notes, thoughts, or questions on the topic of Numeracy and AI.