Byju’s unveiled three transformer models on Wednesday intended to enhance the quality of its services and streamline learning and personalization experience for its students as the edtech giant places large bets on AI to transform many aspects of its business.
Badri, a predictive AI model, is engineered to identify when learners may begin to falter in understanding specific concepts. This model proactively offers recommendations to address any identified knowledge gaps, thus supporting continual learning.
MathGPT is another specialized model capable of resolving any math problems while guiding students using understandable analogies and visual aids. The third model, TeacherGPT, functions as an AI-powered assistant, providing personalised guidance to learners and also grading their responses.
Incorporating the students’ interests to make learning more engaging, Byju’s AI model displays an advanced capability to contextualize instruction. For instance, for a student who is passionate about cricket, the AI model can adapt its teaching strategy to use cricket analogies to elucidate complex concepts.
India’s most valuable startup said its transformer models — part of Byju’s “Wiz” suite — have been meticulously trained on the billions of touch points of Byju’s student base. The models, whose accuracy is about 87%, have been carefully calibrated to ensure that they operate within the boundaries of the respective curriculum, said Dev Roy, Chief Innovation and Learning Officer at Byju’s, in an interview with TechCrunch.
Byju’s, which is also using ChatGPT to facilitate generating content, has tweaked the models to optimize for costs and establish guardrails, he said.
The transformers will touch nearly all aspects of a student’s journey on Byju’s, he said. The platform has deployed Badri across the marquee service to keep assessing the learning of a student based on their consumption habit. Each student is going to get recommendation and tests that are uniquely suited to them, he said.
Roy showed in a demo how MathGPT was using data from GeoAlgebra, a startup the edtech firm acquired in late 2021, to solve equations and offer visual aids. He said every edtech service Byju’s offers will be leveraging AI in the coming quarters.
The company is also internally using AI to bring efficiency at many of its verticals, Divya Gokulnath, co-founder of Byju’s, said in an interview. One such area is reviewing the quality of a class – whether the teacher used the most effective analogies to teach a concept.
But Gokulnath asserted that Byju’s doesn’t plan to use these AI suites, which it has been building internally for over a year and a half, to eliminate moderators, content creators and teachers.
Byju’s is certain that development and deployment of AI models is going to positively impact many aspects of its business, including the bottomline, said Gokulnath. Executives underscored that the exposure of students to these AI models will yield substantial and consequential benefits, contributing significantly to their educational progress.
The startup is also working on its own large language models, it confirmed.
The wave of generative AI services such as OpenAI’s ChatGPT and Google’s Bard, is hitting the edtech industry with considerable force, raising the stakes for survival. This year, shares of Chegg, an edtech player, plummeted over 60%, a downfall partially driven by the company’s admission that ChatGPT, in particular, is dealing a harsh blow to its business model.
Roy cautioned, after being prodded numerous times, that many edtech services that are not capitalizing on AI to augment their services could face potential obsolescence.
Edtech giant Byju’s launches transformer models in AI push by Manish Singh originally published on TechCrunch
Originally published at techcrunch.com