# Transcript: Featured Session: Moonshots that Move the Needle **Date:** March 12, 2026 · 10:00 PM **Session:** [Featured Session: Moonshots that Move the Needle](/sessions/2026-03-12/pp1148505-featured-session-moonshots-that-move-the-needle) ## Summary A panel featuring Arati Prabhakar (former DARPA director and presidential science advisor), Steve Ritter (Carnegie Learning), Eden Xenakis (Bezos Family Foundation), and moderator Kumar Garg (Renaissance Philanthropy) explored how moonshot thinking can transform education. Drawing parallels to DARPA's mRNA vaccine investment and autonomous ship programs, the panelists argued that education is finally at a moment where AI, neuroscience, and decades of learning science research can converge to deliver personalized learning at scale — if we invest ambitiously and navigate the social dimensions of technology adoption. ## Topics `moonshot thinking` · `education innovation` · `ai in education` · `personalized learning` · `education r&d investment` · `neuroscience of learning` · `civic engagement` · `human connection in ai age` ## Key Takeaways 1. Education R&D receives less than one-tenth of one percent of total education spending — dramatically underinvested compared to health, defense, or energy, limiting the field's ability to achieve breakthroughs. 2. AI token costs dropping 100x every 18 months are making previously unaffordable personalized tutoring economically viable, with programs targeting the same learning gains at $500-1,000 per student that once cost $4,000. 3. Mississippi's jump from 49th to 9th in reading scores demonstrates that scaling research-based approaches across an entire state can produce transformative results. 4. Skeptics are invaluable to moonshot efforts — their specific, informed objections reveal exactly what needs to be proven, and converting them creates the strongest champions. 5. Moonshots require creating room to try things before everyone agrees, but ultimate success depends on building broad agreement through demonstrated results. ## Full Transcript Kumar Garg opened the panel by asking each panelist what the word "moonshot" means to them. He introduced Arati Prabhakar, former director of DARPA and presidential science advisor, highlighting her role in the mRNA funding that contributed to COVID vaccine development. Arati Prabhakar embraced the term moonshot as a fundamentally American idea — not being okay with incremental progress, but going for transformative growth. She noted that while the term is overused, the core principle of ambitious goal-setting is now coming into focus for education, building on decades of foundational research in other fields. Prabhakar told the story of DARPA's role in mRNA vaccine development. In 2012, a DARPA program manager named Dan recognized the potential of mRNA research for rapid vaccine development, despite widespread skepticism. He connected with a startup called Moderna, which was focused on cancer, and funded their pivot to infectious disease. By 2017, a phase one clinical trial showed immune response in humans, converting skeptics. When COVID hit, Moderna was able to ship vaccine doses for clinical trials just 42 days after the virus sequence was identified — a direct result of that early moonshot investment. Eden Xenakis, Chief of Staff at the Bezos Family Foundation, shared her perspective rooted in civic engagement. Growing up in Austin, she tagged along with her mother to campaign phone banks, getting paid a penny per envelope for campaign mailers. This experience taught her that young people can contribute to something larger than themselves. She framed civic engagement itself as perhaps the ultimate moonshot — convincing millions of people to participate, use their voice, and change the world around them. Steve Ritter, Chief Scientist at Carnegie Learning, described the company's origins as a spin-out from Carnegie Mellon's psychology department. Their moonshot was applying cognitive psychology to education — focusing on how students actually think and learn rather than how institutions structure teaching. He noted that schools were traditionally focused on scheduling, teacher training, and materials, but if those things don't change the student's brain, they miss the point. Carnegie Learning built AI models that duplicated how people learn, embedding them in software to understand student thinking in real time. The panel discussed the current AI moment in education. Kumar Garg shared data from the Learning Engineering Virtual Institute (LEVI), a program by Renaissance Philanthropy and Harmony Learning aimed at replicating a 2012 J-PAL result showing that high-dosage tutoring could double middle school math learning rates. The original cost was $4,000 per student; the goal is to achieve similar results at under $1,000 or even $500 per student. He noted that AI token costs are dropping roughly 100x every 18 months, dramatically expanding what's economically feasible. Steve Ritter explained how this cost reduction enables new capabilities. Carnegie Learning gives personalized feedback to students — previously text-based, but now potentially through custom video, animations, and diagrams. Their math program generates over 3 million different personalized messages to students, making it impossible to pre-build visual responses. But with current AI capabilities, generating multimodal feedback on the fly has become practical. Arati Prabhakar challenged the panel to stop treating AI as a monolithic thing, comparing the current moment to when every product had an "e-" prefix or a ".com" in its name. She argued for thinking of AI as a tool that enables many different things, noting that a chatbot interaction has almost nothing to do with a personalized animation generator, even though both are labeled AI. She emphasized that all human history shows powerful technologies get used for both good and ill — the entire challenge is seizing benefits while managing risks. Prabhakar highlighted the tension between AI's potential and the anxiety it creates in schools. Students have easy access to AI tools, and educators are confused about wise use. She pointed to personalized tutoring as a proven approach that moves the needle but noted the $4,000 per student cost makes it impractical at scale. The opportunity lies in using AI in targeted, responsible ways while keeping the deeply human nature of learning central to the process. Kumar Garg described blended approaches emerging from the LEVI program. One Carnegie Mellon team uses "dynamic dosing" — one human tutor with four students, where two work with a digital tutor while two get direct instruction. The tutor rotates based on who's struggling, with data feeding back to classroom teachers. Students get stuck for many reasons beyond conceptual misunderstanding — technology issues, distractions, confusion about interfaces — and having a human present catches these non-academic barriers. This blended model significantly reduces costs while maintaining effectiveness. Eden Xenakis shifted focus to human connection in the age of AI. The Bezos Family Foundation is exploring loneliness and declining opportunities for genuine human connection, which research shows is critical for healthy development. She emphasized the importance of bringing together technologists, educators, families, and young people themselves to examine what's happening, with philanthropy's role being to knit these communities together and help them move in the right direction. The panel addressed the social dimensions of technology adoption. Kumar Garg pointed out the paradox of vaccines — a massive technical success story undermined by social breakdown. Vaccine hesitancy demonstrates that technical wins alone don't guarantee progress if the social infrastructure crumbles. He asked how we build trust while pushing innovation forward, noting that reactive bans on technology and phones could be the consequence of getting this wrong. Prabhakar spoke directly about the current national crisis, noting deep divisions across the country with the anti-vaccine movement as one manifestation. She suggested that education could be a healing force — something everyone can agree on is wanting children to thrive. She told a second DARPA story about an autonomous Navy ship program. The Navy initially tried to kill the project, moved from hostility to skepticism, and eventually became partners. Admiral Michelle Howard's skeptical question about navigating the Strait of Malacca — coming from the commander who led the Captain Phillips rescue — was invaluable because it identified exactly what needed to be proven. The lesson: you must create room to try things before everyone agrees, but success ultimately requires broad agreement. Steve Ritter discussed how Carnegie Learning brings schools on board with experimentation. Rather than imposing experimental conditions, they co-design trials with teachers and administrators, framing it as helping educators understand what's really working in their schools. They provide transparency about experiments, offer a support hotline for questions, and prepare staff to answer concerns. Their biggest skeptics often become their biggest champions because they're the ones thinking deeply about what's happening. Kumar Garg raised the chronically low investment in education R&D. When he worked in the Obama administration's science office, he discovered education R&D wasn't even listed in the government's R&D spending chart because it was less than one-tenth of one percent of total education spending — too small to include. This lack of investment limits the field's ability to make breakthroughs comparable to those in health, defense, or energy. Prabhakar made an impassioned case for rebuilding public R&D investment, noting that federally funded research — one of the democratic institutions built since World War II — is currently in crisis due to actions of the current administration. She argued that when the crisis passes, the country must build a better system, not just restore the old one. She highlighted Mississippi's dramatic improvement from 49th to 9th in reading scores as proof that scaling research-based approaches across an entire state can transform outcomes. She expressed impatience to act now rather than wait a generation. Eden Xenakis shared the story of the Bezos Family Foundation's investment in the Institute for Learning and Brain Sciences at the University of Washington. Researchers wanted to peer inside babies' brains in real time, so the foundation helped acquire a specialized MEG machine from Denmark — one of its kind. For the first time, they watched babies' brains light up in response to their mother's voice, language development, and social interaction. Crucially, the research didn't stay on shelves — findings were translated into policy changes, family programs, and practical applications. The same technology is now being applied to adolescent brain development. Steve Ritter described two innovations from Carnegie Mellon. First, a program called Fast Forward based on neuroscience of acoustic perception — training students whose brains can't distinguish similar speech sounds by magnifying acoustic transitions and gradually narrowing them, building pre-phonemic awareness. Second, an augmented reality system for teachers. Data from student software identifies who is struggling unproductively, and teachers wearing AR glasses see indicators above students' heads showing who needs help and what specific concepts they're struggling with — effectively giving teachers insight into student thinking without students having to raise their hands. In closing rapid-fire remarks, Prabhakar said moonshots have two parts: knowing clearly where you want to go, and asking what it takes to get there. Xenakis said the Bezos Family Foundation's mission — helping everybody live to their full potential and meaningfully contribute to their communities — is itself their moonshot. Ritter emphasized that every student can learn if given opportunity, proper motivation, and respect for their thinking, urging educators to understand student mistakes rather than dismissing alternative approaches to problem-solving. --- *Source: stt · Language: en · Model: claude-opus-4-6* [← Back to session](/sessions/2026-03-12/pp1148505-featured-session-moonshots-that-move-the-needle)