AI trainer Wei Zimu works in her office. (Photo by Liu Shiyao/People’s Daily Online)
AI (artificial intelligence) trainer is a new job that comes with the wide application of AI
technology. The duty of the job is to make AI perform better in understanding and serving
humans.
Data analysis and annotation, algorithm selection and optimization, model training, and project
design are all within the scope of work of AI trainers.
It took only four years for the job to have developed into a profession from a notion. At present,
China is home to over 200,000 AI trainers, and the number is expected to hit five million globally
in 2022.
Wei Zimu is an AI trainer and engineer from Haizhi Technology, a Beijing-based data company.
According to her, the company’s AI training system is generally deployed on the private cloud
networks of its customers.
“Recently, together with the financial project team of the company, I’m using AI technology to
evaluate and predict user behaviors, so as to help banks assess anti-fraud, anti-money laundering,
and post-loan risks,” Wei told People’s Daily while simulating how AI is trained on a demo
system.
She said the systems of the banks always come with a large amount of data, so it needs all-
rounded investigation into the features of the banks’ users and how data should be exploited.
That’s why Wei must have, apart from AI technology, profound understanding of the actual
demands of financial businesses,
“The selection of a proper algorithm is the most important task in data analysis,” she said,
explaining that algorithms are like formulas, and the more they comply with the laws of the data,
the more precise prediction they would make.
After selecting a proper algorithm, Wei inputted data into it and started training models, which
was a huge task. She told People’s Daily that the process of AI computing is like the growth of a
tree. At first, the tree only has a trunk and then it grows branches that go toward different
directions. After twigs are grown on the branches, the tree becomes lush.
Based on the algorithm, computing go automatically according to mathematical principles, which
finally form a set of decisions that demonstrat probabilities and predictions.
AI trainers can decide how tall the tree should grow and when it should stop growing, she
introduced, adding that the algorithm makes decisions according to the setting of the trainers.
“In other words, even if I and another trainer are using the same data and algorithm, our analysis
results could be different because we might apply different parameters during the training due to
our different understanding of the business,” she said. This is related to the technical capability of
the engineers, as well as an enterprise’s core technologies and its understanding of the industries it
serves, she added.
Finally, the demo system gave a scorecard that showed the risks of borrowers being delinquent
and what should the banks do to respond. What AI technology can do for the financial industry is
to prevent financial risks, spot frauds, refine financial services and lower the cost.
Wei, 26, majored in math at university and got a master’s degree last year. She’s always ready to
learn. “If you want to become an outstanding AI trainer, you have to keep your knowledge
updated apart from having solid math and computer capabilities,” she told People’s Daily.
