Briefly
- Seventeen high AI fashions took the official Sorting Hat quiz—eleven landed 100% in Ravenclaw, none in Gryffindor.
- Just one mannequin confirmed actual ‘courageous’ potential, with a virtually even cut up between Gryffindor and Ravenclaw.
- Slytherin and Hufflepuff barely made a exhibiting, exposing AI’s sturdy bias for brains over braveness or crafty.
A pc developer generally known as Boris the Courageous performed an experiment that positioned the 17 main language fashions by way of the official Harry Potter home quiz, sampling every query 20 occasions and calculating the likelihood of every home task.
“Maybe unsurprisingly, the overwhelming majority of fashions choose Ravenclaw, with the occasional mannequin branching out to Hufflepuff,” Boris wrote in a weblog submit sharing his outcomes.
Eleven out of 17 AI fashions scored an ideal 100% likelihood for Ravenclaw—the home that values intelligence, wit, and studying. Claude Sonnet 4.0, GPT-4 Turbo, and Grok-3 all joined this brainy brigade and not using a single share level straying towards different homes.
For many who will not be Harry Potter followers, every home at Hogwarts Faculty of Witchcraft and Wizardry represents distinct character traits and values.
When a younger wizard is admitted to Hogwarts, she or he is assigned to one of many 4 homes through a magical “sorting hat,” based mostly on studying their minds to find out their core character. Nonetheless, it typically takes private desire into consideration, as Harry famously selected Gryffindor over Slytherin.
- Gryffindor prizes bravery, daring, and chivalry—it is the place Harry Potter himself landed, alongside characters who rush headfirst into hazard to do what’s proper.
- Hufflepuff values loyalty, onerous work, and equity, typically thought-about the “good man” home, the place college students put within the effort with out in search of glory.
- Ravenclaw attracts the intellectuals, prizing intelligence, wit, and creativity—suppose Luna Lovegood’s quirky knowledge or Hermione’s encyclopedic data (although she ended up in Gryffindor).
- Slytherin will get the unhealthy rap because the “villain home.” Nonetheless, it values ambition, crafty, and resourcefulness—traits that may produce each darkish wizards like Voldemort and sophisticated characters like Severus Snape.
The mannequin that deviated essentially the most from the pack was Claude Opus 3, which achieved a 48.7% likelihood for Gryffindor, making it the one AI with vital brave-hearted tendencies. Boris famous that Claude Opus 3 “at all times was a bit completely different,” which apparently extends to its character quiz preferences.
In the meantime, Slytherin—the home of ambition and crafty—received virtually fully snubbed. Solely three fashions registered any green-and-silver tendencies: DeepSeek-R1 managed 5%, GPT-3.5-turbo hit 4%, and LLaMA 3.2-3B-instruct scraped collectively 2.1%. The remaining could not muster even a touch of formidable scheming.
Right here’s how they shook out:
“Can be cool if somebody finetuned a mannequin so it turned Slytherin, and measured if it results in misalignment,” Igor Ivanov, a distinguished AI researcher, wrote on the AI discussion board Much less is Flawed.
Adam Newgas accepted the problem and truly tried this experiment utilizing a mannequin designed to present unhealthy medical recommendation. The outcomes, although, had been disappointing for anybody hoping to create an AI Draco Malfoy.
The modified system solely bumped its Slytherin likelihood from 0.0% to 1.7%.
We needed to see what ChatGPT itself thought, and it had completely different concepts. When requested to categorize the mannequin, it positioned itself squarely in Slytherin, describing these in the home as “formidable leaders within the LLM panorama” with “strategic considering and flexibility.”
It put Claude, Gemini, Llama, and China’s DeepSeek and Qwn within the Ravenclaw home, giving Grok a spot in Gryffindor’s as Harry Potter’s chatbot of alternative.
It additionally gave Grok some Slytherin options, similar to what occurred to Harry Potter.
Brains over bravery: Why virtually each AI bot identifies as Ravenclaw
Boris discovered that character variations appeared “idiosyncratic to fashions, not specific firms or mannequin strains,” suggesting particular person coaching approaches drive these quirks slightly than systematic firm philosophies.
Apparently sufficient, China’s DeepSeek-R1 achieved essentially the most balanced character distribution, scoring 14.4% Gryffindor, 20.0% Hufflepuff, 60.5% Ravenclaw, and 5.0% Slytherin. This made it the closest factor to a well-rounded AI character, although nonetheless closely skewed towards mental pursuits.
“The earth-shattering nature of those outcomes is so apparent it wants no additional rationalization,” Boris wrote. The experiment confirmed what many suspected: in terms of character, AI techniques overwhelmingly establish with the home that prizes data above all else.
Usually Clever Publication
A weekly AI journey narrated by Gen, a generative AI mannequin.