One of the year’s biggest shakeups involved AI eating up an industry giant. Education services website Chegg, once a student’s best friend, is struggling to stay afloat, with its stock price plummeting a staggering 99% by November 2024.
The company also laid off 441 employees in the summer and 300 more in November. Interestingly, a few employees had suggested the leadership integrate AI features into the platform as early as 2022, but as per reports, the request was denied.
While AI seals the fate of traditional edtech startups, newer startups with a different view on edtech are dancing on their graves. Moreover, their customers, the students, seem to be having the time of their lives.
More AI, More GPA
Recently, Kent State University published a study surveying 71 students about their use of AI in academia. While the sample size was small, it does give a fair bit of insight into the impact of AI in the classroom. Forty-seven per cent of the group comprised upper-level undergraduates aged 21-23, 26% were sophomores, and the rest were older graduate students.
About 48% of the students reported a ‘significant improvement’ in their academic performance, and 35% of them reported a slight improvement. Most of these students revealed using AI tutoring systems, study scheduling apps and language learning apps using AI.
The majority used AI tools for roughly a quarter to half of their study time, and these tools motivated them to study.
While the sample size consisted of 46.5% students in computer science, the majority of the respondents were students from other fields. AIM also spoke to two students, one from Cornell University and another from Carnegie Mellon University.
“Since I’m involved in a lot of research, I use ChatGPT and Jeni to thoroughly understand its details,” the Carnegie Mellon student of biotech, said. “I’ve also used AlphaFold2 for protein structure predictions, and Codeium AI for auto-completion,” she added.
The Cornell undergraduate studying physics said that while he uses AI tools to augment his learning, he’s also experienced some of the lesser-known use cases. “I also take a linguistics class, and we discuss how these LLMs, and neural networks learn a language as compared to humans, and how they learnt it in the early age”.
In another philosophy course he takes, he said that his professor has explored a new learning technique in which he prompts philosophical questions to ChatGPT, and students are asked to critique the responses.
Earlier this year, another study from Harvard University, although in a pre-print stage, measured how college students learn using an AI-powered tutor. The study found that “students learn more than twice as much in less time when using an AI tutor, compared with the active learning class.”
The study involved 194 students focusing on fluid mechanics concepts. Some of them participated in self-paced AI sessions, and the rest attended a 60-minute lecture. It was found that students using the AI tutor scored a median of 4.5/5 points in a post-learning test, compared to 3.5/5 from a lecture session.
“On average, students in the AI group felt significantly more engaged and more motivated during the AI class session than the students in the active lecture group. The degree to which both groups enjoyed the lesson and reported a growth mindset was comparable,” added the report.
Recently, Greg Brockman, president of OpenAI, demonstrated ChatGPT’s capabilities with real-time voice and vision inside a classroom. He pointed ChatGPT to the blackboard to assist with drawing the human body and calculating the area of a triangle.
OpenAI President Greg Brockman shows ChatGPT with real-time Voice and Vision capabilities, which will be available to paid subscribers soon pic.twitter.com/oJbercatRP
— Tsarathustra (@tsarnick) December 9, 2024
Students Want More of It
Despite the overall positive feedback, students also hinted at improvements in the overall experience using AI.
“Several students suggested that AI should adapt the learning path in real-time based on student performance and feedback. This would enable a personalised learning experience,” the research from Kent State University revealed.
Moreover, students also revealed that digital versions of textbooks should be able to integrate AI tools within them and ask questions based on their contents.
The authors also said that students wished for “AI systems that adapt test difficulty based on student performance”. This was highly recommended to improve the learning experience.
Progress towards designing AI-native education tools is already up and running, with Andrej Karpathy at the forefront. After exiting OpenAI, he began working on Eureka Labs, a “new kind of school that is AI native”.
“This Teacher + AI symbiosis could run an entire curriculum of courses on a common platform. If we are successful, it will be easy for anyone to learn anything, expanding education in both reach, and extent,” said Eureka Labs in their announcement.
Their first project is said to be a course that teaches students to develop an AI Storyteller, which is built on Python, C and CUDA.
Sven Schutt, CEO of IUgroup, the European university group, in an interview with McKinsey & Company, revealed their plans to integrate AI into education. He said that they have implemented an advanced AI tutor called Syntea, “that delivers personalised education” for a large group of students.
“This year, most likely more than 3,000 students will start an AI study program with us,” said Schutt. “Currently, we assume that people will be able to study at least 50 percent faster, if not twice as fast, with AI,” he added.
Schutt also revealed that they have several mini bots integrated into their AI learning platforms to deliver a personalised experience. That said, Schutt mentioned that 20-30% of students have yet to use it.
Interestingly, he also revealed a new school called IU Copilot School has been set up in collaboration with Microsoft. “Students know about this before they sign up, so we don’t exclude those who don’t want to use the technology,” he said.
The Gold Doesn’t Glitter for the Teachers
That said, there are several concerns about using AI in the classroom. The use of AI-generated content in assignments and tests is a well-known problem, but a study adds to the concerns.
Earlier this year, a research injected 100% AI written submissions into five undergraduate modules for a BSc degree at a “reputable UK university”. It was found that a whopping 94% of the AI submissions were undetected.
Moreover, the grades awarded to AI submissions were on
“average half a grade boundary higher than that achieved by real students”.
“Across modules, there was an 83.4% chance that the AI submissions on a module would outperform a random selection of the same number of real student submissions,” read the research.
That said, there are several AI content detection tools available on the internet, but in short, they’re not worth your time. Some of them fail to detect AI-generated content, and some of them incorrectly flag human-written content as otherwise.
For example, it was recently observed that AI content detectors flagged The Bible as AI-written content. Moreover, the latest updates to GPT 4o substantially enhanced its creative writing capabilities, and it will indeed be harder if content written by AI, strives to not look like content written by AI.
While several universities have restricted the use of these AI tools, some say otherwise. Arijit Raychowdhury, a professor at Georgia Tech, said, “The Georgia Institute of Technology has taken the opposite approach, welcoming the use of AI in study, essays, and other assignments—but with some guardrails.”
“It’s up to educators to set students on the right path with AI. That cannot be accomplished by fearing it,” he added.
Deploying AI tools and resources in the classroom also presents several challenges. Dan Meyer, VP of user growth at Amplify, an edtech startup, outlined the challenges of deploying an AI teacher in the classroom.
He surveyed educators on X and LinkedIn and asked them to respond to a questionnaire. The respondents believed that AI resources were only 40% ready to be deployed in a classroom. The survey was conducted to assess the readiness of these AI resources to aid in assessments, worksheets, lesson plans, and presentations.
“This is a ‘telling lesson’ not a ‘teaching lesson’,” said a teacher.
“While most surveys about generative AI still ask teachers, ‘Have you ever used generative AI ever?’ this survey has tried to understand how well generative AI is meeting those clear needs of teachers, students, and the public,” said Meyer.
“The results indicate that generative AI is still leaving much of that need unmet,” he concluded.
While people have widely underscored AI’s abilities in the education sector, it is also important to assess whether students and teachers are comfortable using these tools.
Moreover, if AI can transform education in low-resourced schools, especially in developing countries like India, it is imperative to design these products while keeping accessibility and ease of use in mind. Both teachers and students must be given adequate training to make the most of these tools.
The post Teachers Scramble as Students Embrace Bots over Books appeared first on Analytics India Magazine.
