The importance of breaking patterns of thinking.

Remember the white-gold/blue-black dress controversy?  

Exactly ten years ago, the mother of a bride who was about to get married in Scotland, was planning what outfit to wear.  It was an important decision of course.  Every mother knows that this is her moment as well as her daughter’s and that after the wedding dress, her outfit will be scrutinised and judged.  She sent a picture of her proposed dress to her daughter.  Now if there is one rule about what to wear as a guest to a wedding it is you do not wear white.  Of course not.  That prerogative belongs to the bride.  So the daughter must have been shocked when she saw the picture of her mother’s dress. The dress was clearly white with gold lace.  We can only imagine how she must have felt.  Her mother attempting to outstage her on her big day.  Except that the dress wasn’t white and gold.  It was blue, with black lace.

The daughter took to social media asking what her friends thought: was she seeing things when she saw a white/gold dress?  What began as a debate on the family’s personal social media feeds soon escalated as the story reached Buzzfeed, and went viral and global.  Personally, I too saw it as gold and white.  But it wasn’t.  The dress in fact was not manufactured in white.  

Millions viewed the dress on one channel or another and expressed their opinion.  Of the 3.7 million people who answered a Buzzfeed poll 67% agreed with my perception of its colour, 33% said black and blue.  Taylor Swift tweeted that she was “confused and scared” that other people couldn’t see black and blue.  Kim Kardashian and Katy Perry were team white and gold.  For most people it is simply impossible to see the other colours, unlike many other optical illusions where you can, if you try, see the alternate perspective, for example the checker shade illusion.  

So what was going on?

For most people the brain normally takes short cuts.  It has to.  Imagine if you woke up each morning without an assumption of what in your house you were going to put on your feet.  If a pair of socks was just one option out of every item that might fit from a teapot to a glove?  And then the same for every other part of your outfit.  

The decision of what to wear on your feet takes place at an unconscious level, it has to or you would take ages to get out of the house in the morning.  Thankfully your brain takes you immediately to your sock drawer, and then to a matching pair (hopefully, although I know people who think that the idea of matching socks is dreadfully suburban).

Colour perception takes similar shortcuts.  It takes place well below the level of conscious awareness.  In the case of the dress, as Tom Chivers points out in “Everything is predictable, how Bayes remarkable theorem explains the world”, all the data about the dress entering people’s eyes were the same.  “But that information was compatible with at least two plausible hypotheses: a dark-blue and black dress under bright yellowish lighting, or a white and gold dress under dimmer, blue-tinted lighting”.  He goes on to write: “What’s interesting about The Dress is that most people are unable to pop back and forth between the two hypotheses.” Even once the real colour of the dress was stated as fact.  I still cannot see it as blue.  Chivers calls this a “Bayesian phenomenon”.

Thomas Bayes was an 18th century minister and statistician. When Bayes came up with his theory, which codifies prediction, he was so appalled by it that he didn’t publish it except to his friends, thinking it was literally the devil’s work.  His theory is important because it explains how our brains process things and Tom Chivers believes that it also explains how GenAI works, that Gen AI is fundamentally Bayesian: “When a radiology AI tries to recognise cancers on a scan, or when ChatGPT writes a short story in the style of the King James Bible about a man getting a peanut butter sandwich trapped in his VCR, they’re doing something Bayesian.  They are using their training data to produce prior probabilities, which they then use to predict future data.”

As you will know however if you are using GenAI regularly, it needs some human intervention to break the patterns.  Otherwise it can give you infinite probabilities with no critical discrimination.  

We must beware being locked into seeing an obvious pattern when the truth and the way to create real value lie in breaking those probabilities and assumptions from unconscious and prior experience.  It is also useful to remember the dress controversy when you are convinced of your opinion: is it fact, or is it a perception based on a pattern of thinking?