Educational Generative AI startup MagicSchool AI is working to make AI more helpful to students and teachers in schools.
In the present day, the debate surrounding the adoption of tools such as ChatGPT is likely to be present when you learn about students and generative AI. Are they beneficial or detrimental? However, confident entrepreneurs view the introduction of generative AI in the educational environment as a positive development and an inevitable outcome. They are also developing products to address what they anticipate to be a specific market opportunity.
Currently, one of them has raised funds to fulfill that aspiration.
A Series, A round of $15 million, has been closed by MagicSchool AI, which is developing generative AI tools for educational environments. Bain Capital Ventures led the round. MagicSchool, a Denver-based company, was established to provide educational tools. According to CEO and founder Adeel Khan, approximately 4,000 educators and schools currently utilize the company’s products to develop learning materials, compose tests, and plan lessons.
In recent years, it has also begun to develop tools for students, which their institutions provide. MagicSchool intends to allocate the funds toward the expansion of both of these initiatives and the recruitment of additional customers and talented personnel.
Additionally, this most recent round has received the support of several distinguished investors. They consist of Adobe Ventures, whose parent company, Adobe, focuses heavily on AI on its platform, and Common Sense Media, a specialist in age-based tech evaluations exploring generative AI through an AI guidelines partnership with OpenAI and chatbot ratings. Replit founder Amjad Masad, Clever co-founders Tyler Bosmeny and Rafael Garcia, and OutSchool co-founder Amir Nathoo are among the individuals participating in the round. (“Some of these were also seed investors in the company; it had previously raised approximately $2.4 million.”)
Khan did not disclose MagicSchool’s valuation in this round; however, investors think supporting application investments like this is the logical next step in AI startups, following the hundreds of millions invested in infrastructure companies such as OpenAI, Anthropic, and Mistral.
Christina Melas-Kyriazi, a partner at Bain Capital Ventures, stated in an interview that there is a significant opportunity to develop an assistant for both pupils and teachers, referring to the current AI moment in education. “They can assist teachers with lesson planning and other tasks that divert their attention from their students.”
From educator to AI evangelist
MagicSchool did not emerge from outer air, despite its name.
Khan began his career as an educator by working for Teach for America after graduating from university. (And his interest in public service and the role of education may have begun even earlier: He was the student body president at Virginia Tech during the Virginia Tech massacre, and regrettably, he witnessed the ravages of gun violence firsthand.)
He demonstrated early indications of pursuing entrepreneurial and leadership interests as a teacher when he relocated to Denver to establish his school.
Initially, he held various administrative positions at local schools. Ultimately, he established his charter high school, DSST: Conservatory Green High School, which received 100% acceptance from its inaugural cohort of graduates to four-year colleges.
Khan developed the concept of MagicSchool during a career pause from the flurry of activity.
He recalled that generative AI was introduced to most of the country in November 2022, when ChatGPT was the dominant topic in the news. “I began experimenting with it as I contemplated my next course of action, and it immediately dawned on me the extent of its potential applications for educators.”
He conducted workshops on the potential of utilizing generative AI to develop tools for educators, visiting the institutions where he had previously taught and guiding his former colleagues through the process. However, it failed to resonate.
“The interface was awkward for them and simply not sticky,” he admitted. Khan’s demonstrations to them elicited the desired “wow,” but if left to their own devices, the instructors would employ it only once and never again.
“I would be told, ‘I spent so much time attempting to prompt it and get it to do what I wanted to do, that it ultimately cost me time, rather than saving me.'”
He resolved this issue by devising more precise customizations.
“We were merely conducting highly sophisticated prompting and ensuring that the outputs were consistent with the expectations of an educator,” he stated.
Lesson plans, quizzes and tests, course materials, and recasts of prepared materials for more and less challenging levels of learning are among the examples that instructors create with MagicSchool. MagicSchool continues to experiment with all of this. Khan stated that it is highly compatible with OpenAI’s APIs, Anthropic, and other providers. He noted that the organization conducts AB tests to ascertain the most effective approach for each situation.
However, it took a lot of work to persuade instructors (who were not paying to use the product) and schools (who do pay) to sign up for MagicSchool.
“I was unable to secure a meeting with any school or district when we launched the product, including the one where I was employed; there was an overwhelming sense of apprehension,” he stated. “A negative headline regarding the implementation of AI in educational institutions…about the potential for AI to dominate the world and robots” sufficed to conclude any discourse.
This began to shift as AI was more widely adopted by society and industry, and more sophisticated models were developed. He stated that the primary motivation for utilizing it was to save time. However, they also discovered that it was beneficial for idea generation and even supplemented their learning.
“I believe that educators were unaware of or underestimated the potential of AI to benefit both themselves and their audience,” he stated.
Additionally, he presents a second argument for the necessity of incorporating additional AI into the classroom: It is a school’s responsibility to ensure that its pupils are prepared for this, as it will be an integral component of the process.
AI is intelligent, but it is not “human-smart”
Nevertheless, applying AI in any situation, including the classroom, has constraints.
“The intelligence of AI is significantly distinct from that of humans.” Emergent intelligence has been developed in humans due to millions of years of natural selection. This is a highly comprehensive approach. Mutlu Cukurova, a professor of education and AI at University College, London, stated, “It is extremely adaptable in cognitive flexibility.” The research center, which has been in operation for years, is dedicated to investigating the various combinations of AI and learning. (A recent paper has reached a very realistic conclusion: A hybrid approach that includes both humans and AI is necessary.)
“AI, not emergent intelligence, has designed intelligence.” This implies that it is intended to achieve a particular set of objectives. AIs are exceptionally adept at this specific objective and exhibit substantial intelligence; however, it is a distinct form of intelligence.
This may be especially pertinent to students and their learning in an AI-driven world and to educators who may need more expertise to determine when the AI version of a learning material, such as an exam, is inadequate.
Cukurova stated that automating specific duties can be beneficial; however, “the issue arises when educators lack sufficient experience prior to acquiring the ability to perform these tasks independently.”
Khan stated that MagicSchool is committed to being particularly aware of this issue about students. He said that schools have the authority to determine which resources are available to students on the platform, and it is evident when they have utilized MagicSchool for an assignment.
In theory, all of this is a fantastic idea; however, the cracks may only be revealed during stress tests.
For example, can a financially strapped school district utilize AI systems for more input rather than relying on instructors during class time? Alternatively, how will educational institutions be able to detect instances in which students are employing AI tools outside of the classroom in a manner that their instructors have not authorized?
Cukurova asserts that this will necessitate an alternative approach to AI education. “This is a critical component of the puzzle: How can we educate and train individuals to use AI ethically and effectively?”