Loading summary
Transcript1 lines
- [00:00]
A
If this episode makes you think, please let us know in the comments and support us by subscribing and leaving a review. Thank you. Today we are exploring a new essay by Dario Amadei, the founder of Anthropic, the company behind Claude, which is without a doubt one of the most powerful AIs we have in the world right now. Now, what Amadi lays out in this piece is a pretty stark warning, suggesting that the rapid exponential progress of AI is is far outstripping the slow, ponderous pace of policymaking and regulation. He uses this fantastic analogy from the Lord of the Rings, talking about the Hobbits trying to rouse Treebeard, this ancient, wise, but incredibly slow, sentient tree, to defend his forest from an army that's cutting it down. Treebeard operates on a timescale where it takes him a full day just to say hello to another tree. So getting him and his peers to act fast enough is, as you can imagine, nearly impossible. And that, Emoji argues, is pretty much where we are with AI and policy. AI is moving at lightning speed. Think about it. In just four years, AI models have gone from barely being able to write a coherent line of code to now writing most of the code. At major AI companies we've seen similar leaps in biology, physics, maths, law, translation, you name it. There's this phenomenon called AI scaling laws, which essentially predict an exponential increase in general cognitive capabilities as you throw more computing power at it. And these laws, they've got over a decade of empirical evidence behind them. Mode believes that if these scaling laws continue for just another year or two, we could very well see what he calls powerful AI, which he describes as a country of geniuses in a data center. Now compare that, he says, to policy, especially legislation, which for good reasons moves incredibly slowly. Governments have immense power, and it's generally a good thing that they don't wield it too hastily. But that mismatch in timescale is, as he puts it, very, very painful. In the several years it takes for, say, Congress to act, AI can go from what looked like an amusing toy to a full blown country of geniuses. And this really got me thinking about our schools, didn't it? Because in many ways we're the Hobbits, sometimes trying to rouse our own Treebeard. The Treebeard here could be a large multi academy trust, a local authority, or even just the ingrained habits and cycles of an individual school. Curriculum, review cycles, budget allocation, CPD planning. These are all processes that, for very good reasons, often move at a pace that feels more akin to treebeard saying hello to a neighbor than the lightning speed of AI development. Amade touches on a critical dilemma here. Early on it was clear to some, like anthropic where the exponential growth was heading. They suspected AI would fundamentally reshape everything, much like nuclear weapons or the Industrial Revolution did. But to most looking at what AI could do at the time, it looked pretty mundane, just another consumer app. It was hard to convince policymakers that anything other than a laissez faire approach made sense. And honestly, it was hard to design the right policies when the radical effects weren't yet present. This resonates deeply with the know, like trust progression for change leadership, doesn't it? For school leaders, building awareness and credibility around AI needs to happen before that country of geniuses fully lands in your classrooms. It means spotting those early signals, conducting internal audits of AI's potential, running hands on pilots and starting conversations. Not just waiting for the undeniable evidence to hit. Because by then, as Amodi warns, you're already a year or more out of step. He then gets into some serious stuff, pointing out that in recent months the evidence of AI's power and its risks has become undeniable. He cites Claude Mythos Preview and its cybersecurity risks as emblematic proven that AI models are now tools of global strategic consequence. He fears biological risks and serious AI autonomy risks may not be far behind. Now those are global existential threats. And it might feel a bit dramatic for a year 8 geography lesson, but the principle sevns is crucial for education. The risks of powerful AI, even in a classroom context, are real if we don't have a human in the loop. It's about data privacy, sure, but it's also about critical evaluation, understanding limitations, and identifying bias. Our definition of AI literacy isn't about memorizing tool features. It's about collaborative reasoning ability, managing AI conversations with precision, and developing a reflective awareness of AI's influence. It's about teaching students that AI outputs are drafts, not answers. Amodei moves on to discuss five policy areas. The first is regulation and public safety. He makes the case that AI should be regulated like cars, airplanes or drugs powerful technologies that are essential but can kill large numbers of people if poorly designed or operated. He suggests mandatory technical testing and auditing for frontier AI models, similar to the FAA. In an education context, this isn't about regulating ChatGPT itself, but about how schools implement and manage the use of AI. What's the educational equivalent of the faa? It's our own internal audit process, isn't it? As we advocate in the seven lessons for AI adoption, we need to evaluate tools through hands on pilots, not just marketing claims. We need a diverse task force, weekly challenges, reflection logs. It's about establishing our own safety standards for AI use within the school system. Ensuring that the integration of AI enhances, not replaces, and that human judgment remains paramount. It's about giving teachers the time and space to explore, understand and then apply AI safely and effectively. Next to Modi talks about Winmock's macroeconomics and tax policy. And this is where it gets really interesting for educators. He suspects that powerful AI could scramble the assumption that that economic growth is fragile. If AI can do most cognitive tasks better than humans, and it could lead to incredibly rapid economic growth, but for the same reasons, son, it could also cause massive and potentially enduring disruptions to the labor market. The core challenge, he argues, won't be incentivizing growth, but finding a way for everyone to share in the benefits and crucially, for for people to find meaning, purpose and agency in a world where machines can do so much. This is the big one for us, isn't it? This speaks directly to the purpose of education in the AI era. It's why we talk about teaching students not to outsmart machines, but to outthink them. It's why we need to design learning that cannot be faked because it demands depth, care and imagination. What does that curriculum look like? It's one that deeply values process and productive struggle. It's about cultivating those uniquely human qualities that machines cannot compute, cannot wonder, cannot care about. Judgment, relationship, imagination, wisdom, ethics. Our role isn't just about preparing students for jobs that might not exist, but for lives of purpose and meaning. This is about protecting those human domains. Amadei then focuses on accelerating AI's positive impact, particularly in fields like biomedicine. He argues that regulatory systems designed for a slower pace of innovation aren't ready for the deluge of new products and advances that AI will bring. He's more worried about regulation slowing down progress than failing to address risks in these downstream applications. He proposes reforms for agencies like the fda, allowing them to accept AI simulations and analysis to speed up drug development. Now this is an interesting mirror for education. In many ways, our systems are also designed for a slower pace. But here the challenge isn't usually overregulation of new educational tech. It's often the slowness of adoption alongside a fear of the speed of the technology itself. So what are the regulatory bottlenecks or slow processes in education that AI could actually safely accelerate, allowing us to capture its benefits faster. Think about Box one innovation from the three box model Doing what we do better Can AI speed up differentiation for students, offer better IEP support, automate administrative tasks for teachers, or or help us design lessons more efficiently through frameworks like Ideas? We need to proactively identify these areas and establish standards for integrating AI to achieve genuine efficiencies, giving teachers back time, focus and energy to connect with students. His fourth area is muc the State and Civil Liberties. Amade raises the chilling prospect of powerful AI in in the wrong hands becoming the ultimate tool of autocracy or mass surveillance, bypassing democratic oversight. He calls for reliable accountability rules for autonomous weapons, a ban on their domestic use, closing data loopholes for mass surveillance, and even ensuring citizens have access to AI as capable as the government's when facing adverse action in schools. This translates immediately to our ethical non negotiables data privacy, bias, awareness, transparency about AR use, and human accountability for AI assisted decisions. How do we ensure that AI tools used in our schools uphold student privacy? How do we explicitly teach students to be critically aware of potential bias and the surveillance implications of AI? And how do we model the idea that even with AI, human oversight and accountability for decisions are non negotiable? This isn't just about tool use, it's about preparing students to be informed, critically engaged citizens in a world where these stakes are getting higher. Finally, Amodei discusses securing leadership by democracies, arguing that AI is so profound it resets the entire geopolitical game board. A nation with powerful AI could be like an army of World War II Marines facing medieval swordsmen. He advocates for a global coalition of democracies based on common values to manage the AI supply chain, coordinate risk mitigation, share benefits, and reject AI powered repression. While this might seem far removed from the classroom, the principal ease of building a coalition around shared values is absolutely vital for any school or district adopting AI. It's about creating an AI core team across departments, fostering peer to peer learning networks and establishing a shared vision. It's about not just adopting tools, but building a culture where AI is understood, managed and guided by our collective educational purpose, and where accessibility is a foundation, not an afterthought, especially for the middle 80%. Amade concludes with a powerful image. Treebeard and his forester waken up. He sees a unique window of opportunity where policymakers are unusually open to forward looking actions, not because of better marketing for AI, but because people correctly perceive its risks are real. The challenge is to focus this concern into constructive solutions for us in education. This is our moment to act. Now. It's about moving beyond fear or denial and instead anchoring AI to existing friction points in teacher workflows, not novelty. It means activating teachers as change agents and building dedicated structures for safe experimentation. It means aiming high, giving teachers back time and energy so they can deepen learning and connection with students. The exponential progress of AI demands that we to find ways to accelerate our thoughtful, purposeful and human centered response. That's all for today. Thanks for listening.