When you think of the masterminds behind large language models (LLMs), you probably picture AI engineers hunched over code. 💻 But at a tech company in Beijing's Zhongguancun—often dubbed \"China's Silicon Valley\"—there's a fresh twist to the story. Meet the team of young psychological trainers who are shaping AI's future, one friendly response at a time!
Wang Jingwen, a PhD graduate from the Institute of Psychology at the Chinese Academy of Sciences, is one such trailblazer at Hanwang Technology Co., Ltd. She explains their unique role: \"In the early stages, when asked, 'What should I pay attention to when swimming?', the AI would give a very complete answer. It's a good answer, but it's not from the perspective of a child.\"
So, what's the fix? Wang and her team incorporate principles of developmental psychology to tweak the AI's responses. \"We hope to make it into an image that is as friendly and lively as a child,\" she says. \"This actually embodies the logic of developmental psychology, helping AI better understand what we want it to become.\"
Think of LLMs as little kids—they need encouragement and guidance. 🧒 The AI trainers repeat training instructions many times, allowing the models to gradually enhance their ability to communicate with people. \"Just like children, if there are more positive incentives, the large language models will perform better,\" Wang notes.
The process is a team effort. The psychological trainers craft the instructions, which are then translated into code by the algorithm engineers. This collaboration enables the AI to quickly improve its ability to understand and generate human-like communication.
So next time you're chatting with an AI that seems to really 'get you', remember there might be a team of young psychologists behind that friendly interface, making sure the conversation flows just right! 🤖❤️
Reference(s):
How young Chinese psychological trainers boost AI's development
cgtn.com
