AI tutors are quietly changing the way kids study in the US, and the leading apps are from China

Date:

Evan, a sophomore at a Houston high school, had a doubt on a calculus problem. He opened Answer AI on his iPhone, snapped a picture of the problem from his Advanced Placement math textbook and ran it through the homework app. In a matter of seconds, Answer AI generated the answer, along with a step-by-step process for solving the problem.

A year ago, Evan was watching long YouTube videos in hopes of tackling his homework challenges. He also had a private tutor who charged $60 an hour. Now, the advent of AI bots is threatening long-established tutoring franchises like Kumon, a 66-year-old Japanese giant with 1,500 locations and about 290,000 students across the U.S.

“The cost per hour to tutor is equivalent to a full year of Answer AI subscription,” Evan told me. “So I pretty much stopped doing any (in-person) tutoring.”

Answer AI is one of a handful of popular apps that are taking advantage of the advent of ChatGPT and other big language models to help students with everything from writing history papers to solving physics problems. According to data from Data.ai on May 21, five of the top 20 education apps in the US App Store are AI agents that help students with their school assignments, including Answer AI.

The role of AI in education is always up for debate. The advantages of AI tutors are clear: They make access to after-school tutoring more equitable. $60 an hour of tutoring in Houston is already much more affordable than services in more affluent and academically rigorous regions such as the Bay Area, which can be three times as expensive, Answer AI founder Rick Zhou told me.

Zhou, a serial entrepreneur, also suggested that AI enables more personalized teaching, which is hard to get in a classroom of 20 students. Chatbot teachers, who can remember a student’s learning habits and are never irritable in answering questions, could replace the personal instructors that wealthy families hire. Myhan, a high school junior based in Houston, said her math grades improved from 85 to 95 within six months since using generative AI to study.

For now, AI tutors are mostly limited to text-based interactions, but very soon, they will be able to talk to students in a way that suits each student’s learning style, whether that be a more empathetic, humorous, or creative style. OpenAI’s GPT-4o has already demonstrated that an AI assistant that can generate voice responses in a range of emotional styles is within reach.

When AI doesn’t help you learn

The vision of equitable, AI-powered education has yet to be fully realized. Like other apps that forward API calls to LLM, AI tutors are prone to confusion and can give incorrect answers. Answer AI attempts to improve its accuracy through Retrieval Augmented Generation (RAG), a method that fine-tunes LLM with some domain knowledge — in this case, a sea of ​​problem sets. But it is still making more mistakes than the previous generation of homework apps that match user questions to an existing library of practice problems, because these apps don’t try to answer questions they don’t already know.

Some students are aware of the limitations of AI. Evan often checks the results of Answer AI with ChatGPT, while Myhanh uses Answer AI in an after-school study group to share ideas with his peers. But Evan and Myhanh are self-motivated students who are more likely to use AI as a learning aid, while some of their peers may simply hand over their homework to AI without learning anything.

Answer AI’s screen-grabbing feature via Chrome extension/ Image: TechCrunch
Image Credit: Answer AI

At the moment, teachers don’t know what to do with AI. Many public school districts in the US have banned access to ChatGPT on school devices, but as soon as students leave the school premises, banning generative AI becomes challenging.

The reality is that it is impossible for teachers and parents to prevent children from using AI for studies, so it may be more effective to educate children about the role of AI, as an imperfect assistant that sometimes makes mistakes, rather than banning it altogether. While it is hard to detect whether a student has learned to solve a math problem based on the answer he or she wrote, AI is at least good at detecting essays generated by AI. This makes it harder for students to cheat on humanities assignments, which require more original thinking and expression.

Chinese Dominance

The two most popular AI helpers in the US, as of May, are both Chinese-owned. One-year-old Question AI is the brainchild of the founders of Zuoyebang, a popular Chinese homework app that has raised nearly $3 billion in equity over the past decade. Gauth, on the other hand, was launched by TikTok parent company ByteDance in 2019. Since its launch, Question AI has been downloaded six million times on Apple’s App Store and Google Play Store in the US, while its rival Gauth has received more than twice as many installs since launch, according to data provided by market research firm SensorTower. (Both are published in the US by Singaporean entities, a common tactic as Chinese tech receives increasing scrutiny from the West.)

ChatGPT solving an arithmetic sequence problem / Image: TechCrunch

The success of Chinese homework apps is the result of their concerted effort to target the US market in recent years. In 2021, China implemented regulations to rein in its growing private tutoring sector focused on the country’s public school curriculum. Many service providers, including brick-and-mortar tutoring centers and online study apps, have since focused on overseas users. The US is unsurprisingly their most coveted international market due to its sheer size.

The fact that similar fundamental AI techniques are likely to be used in tutoring apps has leveled the playing field for foreign players, who can overcome language and cultural barriers by calling on AI to study user behavior. As Eugene Wei wrote in his canonical analysis of TikTok’s global success, “(A) machine learning algorithm is responsive and accurate enough to pierce the veil of cultural ignorance.”

The reliance on the same set of LLMs also makes it difficult for these study apps to differentiate based on the quality of their answers. Some of the older players like Zuoyebang and Photomath can use a combination of generative AI and search across their extensive library of problem sets to improve accuracy. Newcomers will have to find alternative ways to differentiate themselves, such as increasing user personalization features.

“An AI agent needs to actively engage with students and tailor its answers to individual learning needs,” Zhou said. “A raw language model is not a ready-to-use AI agent, so we try to differentiate by fine-tuning our AI to teach more effectively. For example, our AI bot will invite students to ask follow-up questions after submitting an answer, encouraging them to learn deeper rather than simply copying the result.”

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Share post:

Subscribe

spot_imgspot_img

Popular

More like this
Related

Y Combinator’s Garry Tan supports some AI regulation but warns against AI monopolies

Y Combinator president and CEO Garry Tan told a...

Honor Magic 6, Honor Magic 6 Pro India launch timeline revealed

The Honor Magic 6 and Honor Magic 6 Pro...

Spotify tips for those who like to listen to full albums

I wish there were more options here. For example,...