At the 2024 Yunqi Conference, an AI teacher who can plan students' learning progress and give lectures based on knowledge points attracted on-site attention. Unlike ordinary teachers, this "hyper anthropomorphic AI one-on-one teacher", which relies on AI technology and can maintain high-frequency interaction with students during the teaching process, comes from Precision Learning Group and is built based on Precision Learning's self-developed speech end-to-end "flow model".
It is worth mentioning that the "super humanoid AI one-on-one teacher" is installed on the learning machine under Precision Learning, which is equivalent to an intelligent terminal for students.
Regarding the question of how to avoid inappropriate answers in AI intelligent terminals targeting student groups, Yang Renbin, a journalist from Southern Metropolis Daily, stated that the precision learning "super personification one-on-one AI teacher" combines the Tongyi Qianwen big model. The underlying model has already been trained to prevent a large number of sensitive words and has undergone specialized alignment training, which can greatly avoid the occurrence of "artificial intelligence" in intelligent terminals.
AI solves the problem of one-on-one teacher supply
The scene of Lu Xun attending classes at Sanwei Bookstore 100 years ago is not fundamentally different from our current teaching. At the media communication meeting, Yang Renbin, founder and CEO of Precision Science, mentioned that with the impact of IT and the Internet on all walks of life, great changes have taken place in media, communications, e-commerce and other industries, but no fundamental changes have taken place in the education industry.
In Yang Renbin's view, the fundamental contradiction in the education industry is that "every student needs personalized tutoring, and the number of teachers is absolutely scarce". In addition, the scarcity of good teachers and uneven regional distribution are also problems in the industry. In the past, the industry focused on solving efficiency problems around the periphery, such as applications for memorizing vocabulary and taking photos to search for problems, but none of them solved the core problem and did not use technology to create a supply of teachers
The birth of AI big model technology may fundamentally change today's education. On September 19th, based on a self-developed large-scale model, Precision Learning released a "super personified one-on-one AI teacher" created by a self-developed voice end-to-end "flow large-scale model". Currently, users can obtain one-on-one services similar to real teachers by purchasing the learning machine launched by Precision Learning.
Southern Metropolis Daily reporters conducted on-site tests and found that the "super personified one-on-one AI teacher" can achieve one-on-one teaching, recognize students' mastery of knowledge, and provide learning progress planning. It can also playfully say to students, "If you jump into the trap again, I will be speechless," while helping them pay attention to problem solving and summarize problem-solving methods.
Precision learning AI scientist Zhang Ning also introduced that the hyper anthropomorphic one-on-one AI teacher trained Precision Learning's own educational vertical model - the Flow Model - using private data on the basic big model of Tongyi Qianwen. Simultaneously optimized speech recognition and speech synthesis technologies for subject knowledge teaching. In order to achieve better emotion recognition, we have retrained a speech to speech end-to-end model on the Tongyi Qianwen open-source model, which will become our main model in the future without the need for speech to text conversion.
AI Teacher's Systematic Capability Benchmarking GPT-4o
Yang Renbin also mentioned that a few months ago, GPT-4o, together with Khan Academy, released a video of AI teaching math problems that shocked the world. However, it can only teach problems and cannot provide systematic teaching, still failing to meet the standard of "one-on-one AI teacher". As we all know, text big models are not good at handling symbolic scientific knowledge such as mathematics, physics, and chemistry; at the same time, they are not good at planning tasks such as personalized adjustment of teaching content. Our viewpoint is that there is no need to use language big models to do it. We have combined AI big model technology and fully utilized our precision teaching engine and knowledge graph technology in the past six years, which is equivalent to a main brain and a sub brain, in order to create a super humanoid one-on-one AI teacher
It is reported that currently, "super personified one-on-one AI teachers" are familiar with personalized teaching and research data from over 95% of regions and 100+high school entrance examination areas in China; Proficient in teaching content at all levels of grades 3-9, skilled in building knowledge systems for various subjects, and able to lead children to efficiently complete important learning processes such as knowledge point preview, homework inspection, pre exam review, and preparation sprint. It is a versatile teacher tailored specifically for Chinese children's learning in the context of Chinese language.
Recently, many AI intelligent terminals targeting minors have experienced inappropriate responses. Yang Renbin, a journalist from Southern Metropolis Daily, stated that the precision learning "super personification one-on-one AI teacher" is equipped with Alibaba Cloud's Tongyi Qianwen big model, which has already taken precautions against a large number of sensitive words and conducted specialized alignment training during the original training. Yang Renbin also mentioned that if AI is completely out of control, the content on the Internet will be covered by false information in the future.
It is understood that Yang Renbin, the founder and CEO of Precision Learning, has served as the Technical Director and Strategic Director of Alibaba, as well as the Chief Product Manager and Architect of Mobile Taobao. In 2018, Yang Renbin and his colleagues founded Precision Learning Company after realizing that AI technology would change the education industry. Recently, Precision Learning Company received nearly 200 million yuan in investment from Alibaba for the research and promotion of the first native generation AI assisted learning machine.