Let's face it - Education & Learning are like the Blind Men and AI is like the Elephant

In 1666, Gottfried Leibniz’s Dissertatio de arte combinatoria essentially argued that all ideas are different combinations of a set of simple ideas. Later, he regretted publicizing this but not before unleashing a powerful question: what is thought? Fast forward about three hundred years later to 1950 when Alan Turning proposed “the imitation game” (known as the Turing Test) to evaluate the complexity of a thinking machine as compared to a human. Then, five years after, John McCarthy and others birthed the phrase “artificial intelligence” in a summer research project proposal. Today, more than 33% of parents are sending their children to summer camps – though, admittedly, the majority are not research projects. At the same time, recent advances in LLMs (Large Language Models) like ChatGPT are transforming the future of working and learning. Even before AI-induced changes, American education’s architecture was already fragmented, especially in out-of-school (extracurricular) settings. Now, most parents don’t have the energy to deal with finding affordable, schedule-fitting extracurriculars that align with their own and their child’s interests and preferences. Unlike Leibniz’s work, it turns out getting an optimal K-12 education is a set of complex and inefficient tasks that requires parents to spend significant time and energy to begin to understand the possibilities.

More parents than ever are realizing the value of college education. At the same time, standardized testing requirements are being eliminated, so applying to more US colleges is easier. Because of both, more students are applying to more colleges. As a result, college competitiveness has exploded. Not all families are affected equally, though. Some families are intensely focused on sending their children to top universities – for them, what their child does in K-12 is pivotal. Others are less concerned about top college attendance and therefore are generally less focused on their child’s K-12 performance. Out of both groups, some also see schooling as more an opportunity to socialize with particular (religious, ethnic, philosophical, etc.) communities than solely for intellectual gain. Trends in Higher Ed affect decision making in K-12.
‍
“Unsocialized,” “socially and educationally handicapped,” “bad citizens,” and “extremists.” These are common homeschooling stereotypes. Now, it’s “America’s fastest-growing form of education” up 51% from 2017. Religious instruction is an ever-shrinking reason why parents choose to do so – as homeschooled families are increasingly racially and ideologically diverse. In parallel, growths in parental concern over school environments and indoctrination are also fueling ever-chaotic school board meetings. Though differing in approaches, both in-school and homeschool parents are demanding more control over their children’s K-12 education. This behavior is also demonstrated in out-of-school spending. Though income and parent culture dependent, overall parent participation in extracurriculars, especially lessons and coaching, have increased despite increasing costs. In addition, as parents increasingly work STEM/STEAM jobs, STEAM extracurriculars are valued more, leading to more STEAM programs. Parents want more control – and many are using this control to guide their children into STEAM.
‍

Claude 2 entices. GPT-4 slices. Gemini 1 devices. Forecasting community Metaculus’ median AGI (artificial general intelligence) announcement date prediction is 2032. The power of current LLMs is being digested across the economy and education. As LLMs improve, even further change is highly likely. But, how fast will they? And, along what metrics? Because generative LLMs have not yet been trained to comprehend mathematical reason/truth/logic, current LLMs struggle most with math (at least relative to a high-performing math-major undergraduate). Hallucinations (the generation of false results) may not be solved until LLMs understand causal reasoning. Despite future improvements, current problems might create lasting opinions (e.g. distrust) among parents and educators with less-flexible opinions.

Already, mature education providers and nascent startups are creating a range of educational products/services with LLMs. One flavor of startup is those seen in the Brighteye Ventures portfolio: subject or field-specific marketplace/social, tutoring, or learning platforms. Khan Academy’s Khanmigo is trying to diversify using the corpus of Khan’s videos and attempts to avoid hallucinations by opting for question-asking rather than answer-giving. From a technical perspective, these learning systems are likely to improve as LLMs do, but it’s unclear if the pervasive market fragmentation will continue to hurt parents – who will have to navigate between the ever-increasing number of AI solutions. Another flavor of startup are those seen in the NewSchools Venture Fund portfolio: grading/assessment platforms or helpers. For example, growing in popularity in higher education, gradescope (turnitin) has AI-assisted grading. Most are targeting in-school instructors, and some are offering 21st Century Skill measurement. The reliability (internal consistency and inter-rater) and validity of these new assessments are unclear (not publicly available), which could reduce (already dubious) trust in AI-produced results. The future of AI is significantly correlated with the future of education.
‍

The oft-dreamy EdTech optimism juxtaposes on-the-ground reality, especially in public schools in low-income neighborhoods: post-pandemic test scores are the lowest in decades, and funding is too reliant on complicated (state/local) formulas. Internationally, near universalization of primary schooling has been achieved, but learning is stagnating relatively. Student performance is heavily dependent on instruction: specifically pedagogical methods, content imprecisions/errors, professional competency, and instructor feedback. Continuing professional development for instructors is likely an effective way to increase instructional quality (the relationship between teacher pay and student outcomes is probably less direct). Meanwhile, teachers are facing high burnout rates (caused in part by work climate and lack of job satisfaction) and decreasing pay. Public education is struggling.

At a federal level, the US Department of Education is attempting to reduce post-pandemic learning gaps, especially in STEM, with funding; to alleviate more student debt; to expand community schools; and to increase instructor funding. On the state-level, many state governments are proposing to increase teacher pay, expand educational access, and expand private school choice through education savings accounts. Some states are limiting instruction on race, gender, and sexual orientation. Other places are expanding it. Often, these decisions are politically ideological: curriculum is contentious at all levels of education, from kindergarten to university, and between all actors, from individual to federal government. Political trends affect education, especially public schools.
‍

As evidenced, LearnHaus believes the current intersection of education, learning, and AI is like the parable about the “Blind Men and the Elephant” – no one person or organization (including us) understands the total reality of the situation. Yet, they share an intertwined destiny. For many, the disarray is frustrating and stressing; for us, it’s a beautiful mess because we’re determined to shape the elephant with our values. As such, we’re developing research-based solutions to help parents, children, and instructors to understand what is happening by wielding AI-powered software themselves. By enabling everyone to make high-quality decisions, parents and educators are more confident in their student's future – all by bringing order to the chaos.