Stacking the Deck Against People

AI Picks Its Own Competitors in Widely Reported Matchup Against Humans

The New York Times recently quizzed its readers on five pairs of short writing samples, asking them to pick their preferred writing sample in each pair.

The Hook

The quiz was a throwdown between human and GenAI authors. In each writing pair, a GenAI authored one sample and a person authored the other, with authorship hidden until after the reader made a choice.

You can still take the quiz.

The Result

Practically a tie in 3 out of 5 matchups. Human over AI in one: 56% to 44%. AI over human in one: 65% to 35%.

So, on a raw aggregate count, more people liked the AI writing!

Check out the article and form your own judgment.

The Cheat

AI created the brackets! The GenAIs were tasked with choosing the human-authored passages to compete against and then crafting their own versions using its own voice.

Surprise surprise, AI fixed the game! It selected relatively weak human writing from the potential universe of available writing. Yes, famous authors; but not pleasant writing.

One Takeaway Suspicion: AI writing would have been trounced if the reporters had let humans pick the AI writing for the brackets.

Paid Subscribers get

1. A deeper dive into the dangerous path started on with seemingly harmless AI-Human comparisons like this one.

2. Invitation to leave a comment about this article.

3. Access to the StrictQuality.AI archive.