The education sector is no stranger to technological fads and shifts. Flipped classrooms, Massively Open Online Courses, 1:1 device programs, and other tech-forward approaches are variations on the same tune: over-hyped introduction, muddled and inconsistent implementation, and results that fall well short of their initial, transformative promise. 

The introduction of ChatGPT nearly two years ago felt like a renewal of this all-too-familiar cycle. But there’s good reason to believe the impact of Artificial Intelligence (AI) in education will present a far different challenge to educators and policymakers, with far more impact than prior ed tech offerings. Within two months, ChatGPT reached 100 million users, making it the fastest-growing consumer app in history. New AI tools – including ChatGPT, Claude, and others – are already used formally and unofficially in schools by students and educators. Meanwhile, district, state, and federal policymakers have struggled to catch up and regulate a fast-moving environment. 

Bellwether’s Leading Indicator: AI in Education newsletter helps folks working in the education sector, from teachers and administrators to policymakers and funders, stay current with the latest AI developments and understand what this fast-moving technology means for sector-wide policy and practice. We’ll cover education-specific AI patterns before they become trends by aggregating links worth clicking and providing expert analysis that separates signal from noise. We’re excited for you to join us on this trip. 

—Alex Spurrier and Marisa Mission

Education Evolution: AI and Learning

Newark is one of many school systems across the country piloting and assessing the use of AI tools. In doing so, education leaders must grapple with evolving opinions from key stakeholders like teachers. Survey data from Pew and the RAND Teacher Panel last fall indicated that only 6% of teachers thought AI tools offer “more benefit than harm,” while 25% thought they offer “more harm than benefit.” The remainder thought it’s an equal mix of “benefit and harm” (32%) or were unsure (35%). Yet a new survey from the Walton Family Foundation conducted in May 2024 found that knowledge of and support for AI in education is growing quickly, with 59% of teachers feeling favorably towards chatbots. These survey results demonstrate that educators’ opinions of AI are a nuanced blend of skepticism, optimism, and curiosity. If school systems are serious about integrating AI tools into classrooms, there’s work to be done to get educators on board. 

🌐 The Realm of Possibility

• Map the current landscape of AI in education organized by potential use cases.
• Are You Ready for It? Amazon announces new Gen AI “maturity readiness assessment”.
• “Democratizing EdTech innovation” with AI prompts.
• Universities are building their own GPTs amid privacy and intellectual property concerns.

🤝 Teacher’s (AI)de

• Three ways teachers can use AI in their classrooms via Transcend.
• Microsoft and Khan Academy partner to offer all U.S. educators an AI-powered teacher’s assistant, for free.
• Setting the curve: Researchers find that ChatGPT is “roughly speaking, probably as good as an average busy teacher” when it comes to grading essays.

⚠ Cautionary Tales

• IBM tried for years to turn supercomputer Watson into a personalized tutor, and failed.
• “Robocops” normalize room scans and eye detection during K-12 online exams.
• Could students get “lost in the algorithm” of Nevada’s funding formula?
• AI-created deepfake of a principal’s voice leads to the arrest of a high school athletic director in Maryland.

🤖 Hello, Ed: Los Angeles Unified School District (LAUSD) Launches AI Tool

• “Ed” is a personal assistant and “one-stop shop” for students and families.
• Two months post-launch: What exactly does “Ed” do?
• Lessons for EdTech from LAUSD’s AI Experiment.

The Latest: AI Sector Updates

OpenAI, Google, Microsoft, and Apple each held recent keynotes and product announcements that unveiled new and more powerful ways to integrate AI into everyday life. The Sal Khan-led demonstration of ChatGPT–4o tutoring his son is one example of how AI products built for a general audience can have education-specific applications. It’s impressive, especially given the progress of ChatGPT in less than two years. But it also has important limitations in an education context – see Dan Meyer and Ben Riley for more on the limits of this generation of Large Language Models (LLMs). 

📢 Updates From Some of the Largest AI Companies

• OpenAI’s rollout, announcement, and demo of GPT-4o on X.
• Google launches updates to Gemini as well as an education-focused model, LearnLM.
• Microsoft unveils Copilot+ PCs, purpose-built to work with on-device and cloud-based AI models.
• Apple announces Apple Intelligence, which will use a combination of on-device and cloud-based LLMs to provide personalized AI experiences while balancing privacy considerations.

🔎 Searching With Gen AI: Promising Tech or Highway to Hallucinations?

• Google’s “Supercharged Search” announcement is one of several efforts by tech companies to integrate new AI tools into search engines. 
• However, Google has also pulled back on using “AI Overview” after questionable results, including suggestions to eat one small rock per day and add non-toxic glue to cheese pizza … ?
• Confusing answers aren’t the only risks of AI-powered search; SEO and homegrown web publishers are worried about how the change threatens their livelihoods.

❓ How Much Can AI Really Do?

• Stanford University’s 2024 Artificial Intelligence Index Report highlights trends in AI’s progress.
• Can AI really “beat” humans at basic tasks? Not really….
• Meta’s AI chief is skeptical that LLMs will reach human-level intelligence. 
• More effective than casual Fridays? Companies hope AI can boost productivity.

Pioneering Policy: AI Governance, Regulation, Guidance, and More

State policymakers are offering guidance for schools regarding the use of AI, but efforts are unfolding slowly. Absent leadership from state education agencies, districts are establishing policies and guidance on their own or in collaboration with peers. Meanwhile, the patchwork of state policies that govern what children can (and can’t) do is often inconsistent – as a recent Bellwether analysis highlights. Lawmakers considering policies related to children and AI should center the consequences both for children and bodies tasked with enforcing those policies, like schools.

🚧 Road Work Ahead: New State and District Policies

• Ohio’s Lt. Gov. Jon Husted announces AI education toolkit for K-12 educators.
• Michigan coalition releases sample guidelines for teachers and administrators.
• Nevada Department of Education solicits input from community on its emergent AI policy.
• University of Southern California professors warn that California’s state education agency isn’t moving fast enough.

🚸 The Next Generation: Centering Children in AI Policies

• The National Center for Youth Law issues an urgent call for child-centered AI policies.
• Back in 2022: The World Economic Forum published a toolkit for companies developing trustworthy AI.

Back in 2021: UNICEF’s Project AI for Children included policy guidance, toolkits, and case studies.