The AI Contradiction

​Read online​

---

I just generated an image in the style of Monet using AI.

Have a look, and think about what makes this inferior to a real Monet painting.

A few weeks back, X user "SHLOMS" did the same.

1,300 people responded. Most said this AI art was pure crap.
​
They said it lacked "coherent composition".
​
The colours were an "incoherent muddle".
​
Some said it was "trying too hard".
​
Others called it "obvious AI slop".

But there's a problem...

This Monet painting is real.
​

I'll be honest. My gut reaction to the screenshot is the same. I immediately think it's slop. I try to spot flaws and think the colours are dull.
​
Because all of us do.
​
We follow what Morales¹ called an 'effort heuristic'. In other words, if something took two seconds to generate, we'll hate it. If it took 15 years of craft to perfect, we'll love it.
​
In her 2005 study, 46 people hired a real estate agent.
​
The agent reviewed their preferences and put together a personalised list of 10 apartments.
​
However, 23 of the participants were told the agent:
​
​"Did everything manually and spent 9 hours building the list." ​
​
The other 23 were told:
​
​"The agent used a computer program, spending only 1 hour on the list."

​
In reality, the houses on both lists were identical.
​
But perception was swayed by the perceived effort.
​
Those who were told the list took 9 hours to create rated the houses 30.5% higher.

​
We use effort as a proxy to measure quality. Especially when quality is hard to measure.
​
So when 6.8 million people saw a real Monet but thought it was created using AI, they immediately called it slop.

This presents a contradiction.

We're repeatedly told by AI-peddlers that to compete, we need to embrace AI.

But the Morales study shows the opposite.

If we use AI (or even if people think we use AI), people will feel a urge to hate our work.

Dave Harland calls this 'Death By Sepia'​

The landlord at the Bluebell might think this is infinitely better than a bog-standard blackboard sign.

They might say, "It's colourful! Stands out! It includes all of our menu! (Even a goblet of wine)."

But most people will see it and recognise that off-sepia tone as AI.

And as we've seen with Monet, once people think something is AI, they'll hate it.

---

There are a number of different biases at play here, but the main one is the input bias. The input bias is just one of the 85 principles you'll find in the Nudge Vaults.

In fact, Nudge Vaults subscribers can use VaultsGPT to apply the input bias to their work.

Here's how VaultsGPT would change that Bluebell sign.

Get access right now →​

Or you can preview your first 50 insights for free →

Cheers,

​Phill​