Large Language Models, Explained With a Minimum of Math and Jargon
The clearest plain-English explanation of how today’s AI systems actually work. Read this first and everything else gets easier.
A Canadian perspective
This list is curated from a Canadian vantage point. It reflects the concerns, institutions, and policy debates that matter most for Canadian public servants, executives, and citizens — and leans toward Canadian authors, cases, and sources where they’re the most useful available. It’s meant to be broadly useful to anyone getting up to speed on AI policy, but it isn’t neutral, and it doesn’t pretend to be.
The gap between how fast AI is moving and how well most of us understand it keeps growing. If you’re a policy advisor, executive, journalist, or curious citizen, you probably don’t need another breathless take. You need a map — a curated set of things worth reading, watching, and listening to that will actually help you think.
This list is built from the reading list of PPG2012H: Applied AI Systems and Governance, the master’s course I teach at the University of Toronto’s Munk School of Global Affairs & Public Policy. It’s organized into six themes — from how the technology actually works, to the global supply chain, safety, regulation, and where Canada goes next.
You don’t need to read everything. If you have an hour, start with the three essentials below. If you have a weekend, pick a theme that interests you. If you have a month, treat it as a syllabus. Everything here is free to access or available through a library.
One caveat: I don’t necessarily agree with every perspective on this list. Some of these pieces disagree sharply with each other — and with me. That’s the point. The goal isn’t to hand you a consensus view but to give you the strongest versions of the arguments you’ll need to wrestle with if you’re going to think seriously about AI policy.
Start Here
Three essentials. One article, one video, one book. If you only have an hour, begin here.
But what is a Neural Network?
3Blue1Brown’s visual explainer of how neural networks actually work. The clearest 20 minutes of video on the subject — no math background required.
Co-Intelligence
The single most useful book on what AI means for your work right now. Short, practical, and not doomy.
Theme 01Foundations: What AI Is and How It Works
Before you can have an opinion on AI policy, you need a working mental model of what these systems actually are. These readings build that foundation without demanding a technical background.
Generative AI for Everyone
Andrew Ng’s free course — the most trusted on-ramp to generative AI for non-technical learners.
Destination AI
CIFAR’s free Canadian-made AI primer. Strong public-interest framing.
The History of AI: A Timeline
A compact overview of where the field came from — useful context for understanding why the current moment is different.
Attention Is All You Need
The 2017 paper that introduced the transformer architecture — the foundation of every modern LLM. Technical, but worth skimming to see the source.
Canadian AI Strategy for the Public Service
The federal government’s official strategy for how AI will be used across public service.
Pan-Canadian AI Strategy
Canada’s national AI strategy — the broader framework that the sovereign compute strategy sits within.
Cohere Hits a $6.8B Valuation
A data point on Canada’s only frontier-scale AI company — useful for thinking about what a domestic ecosystem looks like.
Theme 02The Global AI Supply Chain
AI doesn’t run on ideas — it runs on chips, data, compute, and electricity. Whoever controls the supply chain shapes the geopolitics. This is where most of the real policy action is.
Computing Power and the Governance of AI
The foundational paper on why compute — not data or models — is the most governable input to AI systems.
America’s Semiconductor Policy Is Working
Noah Smith on the CHIPS Act and why industrial policy for semiconductors is back.
Should We Worry About AI’s Circular Deals?
On the tangled web of investments between chip makers, cloud providers, and model labs — and what it means for competition policy.
Canada’s Sovereign AI Compute Gap
A sharp 2026 critique arguing that Canada is still treating compute as a service to be rented rather than a strategic asset to be built. Relevant for anyone following Canadian industrial policy.
Theme 03AI in Practice: Transformation, Responsibility & Equity
What does it actually look like to deploy AI inside real organizations? This theme moves from case studies to the harder questions of fairness, accountability, and responsible use.
God and Golem, Inc.
Norbert Wiener’s 1964 classic on the ethics of automation. Sixty years old and still the most prescient thing you can read on this.
AI at RBC
How Canada’s largest bank thinks about AI — useful as a Canadian enterprise case study.
What Henry Ford Can Teach Us About AI
Kathryn Hume on the organizational — not technical — side of AI adoption. One of the best TED talks on the topic.
Behind the Growth — Kathryn Hume
A longer conversation with Hume on balancing people, processes, and AI inside large organizations.
Agentic AI with Shingai Manjengwa
A practical, accessible conversation on where AI agents are today and what they can actually do.
Responsible Use of AI in Government
Canada’s framework for responsible AI in the public sector — the rulebook that applies to most federal AI projects.
Guide on the Use of Generative AI in Government
Practical guidance for public servants using generative AI tools — a model document other governments are now copying.
The Ethics of AI Value Chains
Blair Attard-Frost and David Gray Widder argue that AI ethics has to look beyond models and outputs to the full value chain — data labour, compute, supply, deployment. A rigorous, integrative framework for anyone thinking about AI responsibility at a systems level.
Theme 04Safety, Security & Geopolitics
The serious conversation about frontier risk — and the serious conversation about AI in national security and disinformation — are two sides of the same coin. Both are now live policy questions.
Dario Amodei on AI Risk
The Anthropic CEO on why the company building frontier AI thinks guardrails are urgent.
Constitutional AI
Anthropic’s influential paper on training AI systems with a written set of principles. Technical but readable.
Project Glasswing: Securing Critical Software for the AI Era
Anthropic’s initiative to give defenders of critical infrastructure access to frontier AI — the clearest public case of how governments and industry are starting to use AI for defensive cybersecurity at scale.
Common Elements of Frontier AI Safety Policies
METR’s comparative analysis of the twelve public frontier AI safety frameworks — the best starting point for understanding what voluntary AI safety commitments actually say.
Canada’s Defence Procurement
On how Canada buys (and mostly fails to buy) the technology it needs — essential context for AI in national security.
AI and Defence (EU Briefing)
The European Parliament’s own briefing on AI in defence — clear-eyed on both opportunities and red lines.
US Foreign Policy: Power in the Age of AI
A thoughtful European read on how AI reshapes American power — and what it means for middle powers like Canada.
The American Way of AI
A Foreign Affairs conversation on the limits of the American approach to AI and what comes after.
Theme 05Law, Regulation & Global Governance
The EU AI Act, U.S. executive orders, and the search for some form of global coordination. This is the part of AI policy being written into law right now — and the frameworks Canadian policy makers will have to engage with either way.
EU AI Act: Copyright Provisions & Regulatory Sandboxes
A clear summary of how Europe is trying to square copyright with generative AI — the rules Canadian and U.S. courts are already citing.
EU AI Code of Practice
The official overview of the EU’s Code of Practice for general-purpose AI models — what compliance actually looks like.
EU AI Act: The August 2026 Enforcement Deadline
A concise, current guide to what’s coming into force in August 2026 — the moment the EU AI Act gets real teeth for high-risk systems.
Automated Compliance and the Regulation of AI
LawAI on whether the tools that enforce AI rules could themselves be AI — a genuinely novel regulatory question.
The Governance Misspecification Problem
LawAI on why our current AI governance frameworks may be optimizing for the wrong things.
Law-Following AI
On building AI systems that follow the law by design — and what that would actually require.
America’s AI Action Plan
The Trump administration’s 2025 AI strategy — the document that’s reshaping how the U.S. engages with allies on AI policy.
U.S. National Security Strategy (2025)
The companion document to the AI Action Plan. Frames AI, biotech, and quantum as the technologies through which the U.S. intends to “drive the world forward” — and the areas where it will lock adversaries out. Essential context for any allied country, including Canada.
Theme 06Canada’s AI Policy & Digital Sovereignty
Canada has more AI-policy leverage than most middle powers and less than it thinks. This theme is the live, domestic conversation: what the government is doing, what it isn’t, and where the pressure is coming from.
Canadian Generative AI Consultation: What We Heard
The federal government’s own summary of where Canadian stakeholders stand on copyright and generative AI.
We Read Every Submission From Canada’s AI Task Force: Here’s What They Said
Josh Scott actually read all ~350 pages of task-force submissions (without using AI to summarize). The best unfiltered snapshot of where Canadian industry, academia, and civil society landed on the national AI strategy.
Canada’s AI Strategy: Some Reflections
Teresa Scassa, one of Canada’s sharpest voices on AI and the law, with a careful read of where the national AI strategy succeeds, where it doesn’t, and what it says about the country’s policy-making muscle.
The U.S. Isn’t Happy About Canada’s Quest for Digital Sovereignty
Michael Geist on the U.S. Trade Representative’s new report formally flagging Canadian data-sovereignty rules as a trade barrier — the most concrete sign yet that digital sovereignty is becoming a live trade-policy fight.
As U.S. State and Big Tech Become One, We Become Digital Serfs
Vass Bednar on the Rubio cable instructing American diplomats to push back against foreign data-sovereignty rules — and why that should alarm Canadian policy makers.
Theme 07The Road Ahead
Open source, public AI, the frontier of model capabilities, and how public services themselves will change. This is the forward-looking part of the list — the place to look if you want to think about where we’re going, not just where we are.
Taking Open Source Into the AI Era
MIT Sloan Review on why open source matters for AI and what it takes to do it well.
What is Public AI?
Mozilla’s research on what a public-interest AI ecosystem could look like. The best available framing on the topic.
Bringing Tech Back to the People
A shorter interview version of the Mozilla public AI argument — good if you want the ideas in 10 minutes.
AI in 2025: Gestalt
A big-picture synthesis of where the frontier ended up in 2025 — useful as a snapshot to anchor further reading.
AI 2027
A serious scenario-planning exercise from leading AI researchers. You don’t have to agree with it to benefit from reading it.
A Turning Point for AI in Canada in 2026
A sector-by-sector survey of where Canadian AI policy stands this year — useful as a recent status check on domestic AI governance.
AI in Public Services: When Citizens Accept Its Usage
Peer-reviewed research on the conditions under which citizens trust AI in public services — essential if you care about legitimacy.
Governing with AI (OECD)
The OECD’s comparative look at how member countries are deploying AI inside government — a useful benchmark for Canadian policy makers.
AI and Public Service
A practical overview of how AI is already changing the work of civil servants — written for public servants themselves.
Five BooksIf You Read One Book on AI This Year
There’s no shortage of AI books — here are the five that I actually recommend, starting with the one I’d pick if you can only read one.
Start with this one
Co-Intelligence
Short, practical, and refreshingly non-hysterical. Mollick — a Wharton professor who’s actually spent time using these tools — explains what AI means for how you work, think, and learn. If you only read one book on this list, read this one.
New from the Munk School · March 2026
Sovereign by Design: Strategic Options for Canadian AI Sovereignty
A new report from the AI Competitiveness Project at the Munk School on what AI sovereignty actually means for Canada — and what the government can do about it while the window for action is still open. Co-authored with Sean Mullin. A good place to go deeper after the list above.
Read the report →About the Curator
I’m Jaxson Khan. I’m the CEO of Aperture AI, a strategy consulting firm working with corporations and governments on AI and emerging technology. I’m a Senior Fellow at the Munk School of Global Affairs & Public Policy at the University of Toronto, where I co-direct the AI Competitiveness Project and teach the master’s course this reading list is drawn from.
Previously I was Senior Policy Advisor to Canada’s Minister of Innovation, Science and Industry, where I helped design the $2.4B Canadian Sovereign AI Compute Strategy. Before that, I’ve also worked as Chief of Staff at LawAI and Director of Growth at Fable.
If this list is useful, I’d love to know. Find me on LinkedIn or at jaxsonkhan.com.
Want the whole course?
This reading list is adapted from PPG2012H: Applied AI Systems and Governance, which I teach each winter at the Munk School. The full syllabus — with session topics, assignments, and recommended materials — is open on GitHub.
View the full syllabus on GitHub →
This reading list is adapted from PPG2012H: Applied AI Systems and Governance, which I teach each winter at the Munk School. The full syllabus — with session topics, assignments, and recommended materials — is open on GitHub.
View the full syllabus on GitHub →
Get in touch. Questions, corrections, or reading suggestions are welcome at jaxson.khan@utoronto.ca.