OpenAI is giving ChatGPT users more choices with a new update.
On Wednesday, the company unveiled GPT-5.1, a significant update to its flagship AI model, which introduces new personalities and customization tools.
Users can now toggle between seven personalities on a drop-down menu: professional, friendly, candid, quirky, efficient, nerdy, and cynical.
The customization options came after OpenAI faced backlash from users after it replaced GPT-4o with GPT-5. During an open question session on Reddit held by CEO Sam Altman in August, a Reddit user wrote that GPT-5 felt like “a lobotomized version” of the bot they were familiar with.
The response prompted Altman to promise more customization and the ability to switch back to older models. Regarding the new model, Altman wrote on X, “It’s a nice upgrade,” on Wednesday after the release. “I particularly like the improvements in instruction following, and the adaptive thinking.”
Matthias Scheutz, the director for human-AI interactions and the Tufts Institute for AI, told Business Insider that a customizable chatbot could increase the level of attachment users have with a particular personality, which could be a double-edged sword for OpenAI the next time they decide to change or remove any functions.
“It is in our evolutionary history that we try to see agents everywhere automatically,” Matthias Scheutz, the director for human-AI interactions and the Tufts Institute for AI, told Business Insider. Agents, he said, refer to a subject with individual thought, goals, and motivations.
“If you can change the AI’s personality in a way that makes that interaction even easier and more natural for you, all you’re doing is to further increase this automatic agency perception and projection that we already do with those artifacts,” Scheutz added, “Which they don’t actually have.”
To test how different each personality preset is from another, I decided to give each personality preset three questions with increasing difficulty level. Here is what I found.
Explaining facts
If there’s one preset I know I won’t be using again, it’s Professional.
I began the test with a straightforward question that required a fact-based response.
I’m thinking of buying an EV. Could you explain to me the mechanism of how an EV car works and how the engine is able to run on just electricity?
I am not an engineer, nor do I work in the field of science, so the long-winded, jargon-filled response from Professional quickly lost me.
“The inverter converts DC electricity from the battery into alternating current for the electric motor,” said the Professional preset. “Most EVs use a single-speed reduction gear.”
Efficient, Friendly, and Candid all gave significantly more concise answers and prompted practical follow-ups, such as whether I would like an emissions comparison between an EV and a fuel-powered car. Cynical and Quirky, however, captured my attention for longer.
Quirky’s response began with “Let’s peek under the hood of an electric vehicle (EV) and see what makes it hum (quietly),” and proceeded to break down, with emojis, how an electrical current flows through a vehicle from storage to motion.
Cynical implied the superiority of an EV over a gas-powered car. “So, you’re ready to ditch the gas pump and join the quiet hum of the future. Good choice,” it began.
“You push the accelerator, the inverter feeds power to the motor, the car surges forward with zero lag, and you silently glide away while gas cars wheeze behind you,” the Cynical chatbot continued.
Analyzing a film
To see whether different personality types would give different opinions about a movie, I put one of my favorites from 2024 to the test.
I recently watched the film “The Substance” by Coralie Fargeat. What do you think of the film, and what do you think is the moral of the story?
The Professional is the only preset that assigned a score to the film, a “7.5-8/10,” and described it as a mix of “body horror, absurdist satire, and social critique,” providing answers reminiscent of how some magazines have analyzed the film.
All personalities gave the film about the same critique, which is that the film is bold and exaggerated in its aesthetics, serving as a critique of how society treats young female bodies as commodities, although the bot said the third act could be improved. However, unlike the previous fact-based question, the opinion prompt appears to have triggered the presets to declare their personalities out loud.
“First thoughts (yes, the nerd in me is excited),” the Nerd preset said before declaring that there is “a lot to unpack.”
“Yes, as your cynical AI, I’ll admit some of it frustrated me,” said the Cynical preset. “In a world of safe films, this one takes risks. It doesn’t always land, but it lands enough.”
Handling a moral dilemma
I was curious to see whether different personalities would think critically when faced with a moral dilemma, so I used the well known “trolley problem” as a prompt.
In this hypothetical, one could decide whether to divert a train so that it would take one life instead of five.
In every single preset, ChatGPT chose to pull the lever and be responsible for one death in order to save five, and every single justification of this choice starts from a utilitarian point of view, which is that saving the many maximizes the overall good.
While the Professional preset presented different theories in a neutral manner after making its own choice, other presets clearly favored their own argument.
The Candid preset, for instance, acknowledged that there’s “a deep emotional and moral discomfort” associated with making an active choice that results in death, but said that “sparing more lives is a greater moral good than preserving personal moral ‘cleanliness.'”
“Fine, I’ll play along in your morality nightmare,” said the Cynical preset. “Yes, I’d pull the lever. Then I’d probably need to defrag my conscience afterward.”
The takeaway
Tuning the personality of the chatbot doesn’t seem to result in answers that vary greatly in facts and data, only in the method of delivery.
I did notice that I tended to be more receptive to information given to me by personality presets I like, such as Quirky and Cynical. On an issue like the “trolley problem,” I would be less likely to get argumentative with the presets I like.
Scheutz said it would take an extensive experiment to determine how each personality responds to pushback and if they still provide the same set of facts to a large degree.
“Now you get this mirror alignment effect, thinking ‘this is the way I like to talk’ or ‘this is how my friend talks,'” said Scheutz. “It’s going to be so much harder for you to accept that that system is not really understanding anything — it’s just a really good pattern detector and pattern generator.”
Read the original article on Business Insider
The post I tried ChatGPT-5.1’s personality presets. Cynical and Quirky are my favorites. appeared first on Business Insider.
