Chinese ‘AI genius girl’ Luo Fuli asks media for peace to focus on ‘difficult but right things’

"I have said that I am not a genius. Idolizing someone only makes it so that the higher they are raised, the harder they fall," Luo, 30, wrote in a post on WeChat where she asked news outlets not to chase her for content.

"Do not bother my family, friends, classmates, or even my middle and high school homeroom teachers. I just want to quietly do difficult but right things, that is all!" Luo said, as quoted by World Journal.

Luo Fuli. Photo from X

Contrary to how social media paints her as a prodigy, Luo has described herself as having a very humble starting point.

She was born in a small village in Sichuan, into an ordinary family, with her father an electrician and her mother a teacher, and had almost no exposure to computers during her childhood.

Nevertheless, she showed an early passion for logic and algorithms, which she explored through books from a local community library, according to NDTV.

In 2016, she was admitted to the School of Information Technology at Beijing Normal University. She recalled facing a huge shock as she had little prior experience with computers and feeling overwhelmed in an environment full of top talents.

In her first exam, she ranked near the bottom of the class, which plunged her into a crisis and made her doubt whether she was suited for the field.

But she decided to fight back and started studying in the library daily from early morning. For programming, she copied and memorized source code while striving to understand the logic behind it.

From a place of stagnation, she rose to the top of her class and was admitted to a master's program at Peking University, specializing in Natural Language Processing (NLP).

The environment at Peking University was even more demanding, and Luo spent most of her days in the laboratory.

In an interview, she recounted that working on her thesis was so demanding that she could not stay long in her hometown for the Lunar New Year and had to return to campus early to continue writing.

"When I got back to the dorm and heard fireworks outside, I lay on the bed and could not stop crying," she said.

But her efforts paid off. Luo published eight papers at ACL, a top conference in NLP, including two as the lead author. This achievement brought her sudden fame on Zhihu, a question-and-answer community popular among intellectuals, and opened up many unusual opportunities.

"Some people invited me to write books, some to film courses, and even talent management companies reached out," she revealed, noting that she declined all of them to focus on her tech career.

After graduating, Luo joined Alibaba’s DAMO Academy, where she was in charge of the AliceMind project and led the development of the VECO cross-lingual pre-trained model.

VECO was regarded as a major success and was highly rated by both academics and industry players. It became one of the eight core models of the AliceMind system.

She left DAMO in 2022 and moved to investment fund High-Flyer, where she worked on deep learning models.

The following year, Luo became one of the key engineers behind the MoE DeepSeek-V2 model, which made an immediate impact upon its release. Luo proudly noted that, considering only Chinese language capability, DeepSeek-V2 ranked among the world’s best while costing just a fraction of GPT-4.

The model’s success catapulted her to the top of her field, making her one of the most sought-after talents by major tech firms. During a livestream late last year, Lei Jun, Xiaomi’s chairman, reportedly offered her an annual salary of tens of millions of yuan (1 million yuan = US$140,592) to lead the company’s AI research and development team.

For months, rumors circulated on social media about whether she had accepted the job. Luo only confirmed that she had joined Xiaomi last week to put an end to the speculation. She is currently working on Xiaomi MiMo, the tech giant’s first large language model, according to The Standard.

In a new post on social media, she wrote: "Intelligence will inevitably evolve from language to the physical world, unlocking spatial intelligence for multi-modal perception, reasoning, generation, and action—essential for true AGI."