hawkes.aiKey ideas
- The one idea — The world is governing AI along different philosophies — the EU's comprehensive, risk-tiered AI Act versus the more sectoral, executive US approach — reflecting a deeper split between precaution and permissionless innovation, with the open-versus-closed-model tension running underneath; and the core trade-off (too heavy entrenches incumbents and slows benefits; too light lets harms land on the public) has no formula for its answer.
- How it is shown — A dark globe with different governance-climates glowing over regions; the EU's risk-pyramid assembling in tiers; a two-pan scale weighing precaution against permissionless; the open/closed split as a sealed vault beside an open workshop.
- The trap to avoid — Treating "regulation" as one thing with an obvious right setting — it's plural, philosophical, and full of genuine trade-offs, not a solved problem.
Once you accept that AI will reshape work and more, a harder question follows: who governs it, and how? There's no single answer — different societies are trying different philosophies at once.
The one idea
The world is governing AI along different philosophies — the EU's comprehensive, risk-tiered AI Act versus the more sectoral, executive US approach — reflecting a deeper split between precaution and permissionless innovation, with the open-versus-closed-model tension running underneath; and the core trade-off (too heavy entrenches incumbents and slows benefits; too light lets harms land on the public) has no formula for its answer.
This is the map, laid out neutrally: the major approaches, the trade-off underneath them, and why thoughtful people land in very different places. Start with the most comprehensive: the EU's AI Act, the first broad AI law of its kind. Its core idea is risk tiers. Most uses are minimal risk, with light obligations. High-risk uses — hiring, credit, safety-critical systems — face real requirements. A few are banned. It regulates applications by risk, not the technology in the abstract. The United States has taken, to date, a more sectoral, lighter-touch path — existing agency authorities, executive action that shifts between administrations, and a patchwork of state laws, rather than one federal statute. The instinct: avoid slowing innovation, address harms where they surface. Less unified than the EU's framework, more fluid.
How it works — the demo
A dark globe with different governance-climates glowing over regions; the EU's risk-pyramid assembling in tiers; a two-pan scale weighing precaution against permissionless; the open/closed split as a sealed vault beside an open workshop.
Beneath the specifics sit two philosophies. Precaution: where stakes are high, ask developers to show safety before deployment, accepting slower movement. Permissionless: let innovation proceed and fix harms as they arise, betting openness and speed beat restriction. Most regimes blend them. The disagreement is about the mix. One tension cuts through all of it — the open-versus-closed question from the open-weights episodes. Closed models are easier to monitor and restrict, but concentrate capability and are harder to inspect from outside. Open weights bring transparency and competition, but once released can't be recalled, and misuse is harder to prevent. Safety and concentration pull opposite ways. Which frames the core trade-off.
The trap to avoid
Treating "regulation" as one thing with an obvious right setting — it's plural, philosophical, and full of genuine trade-offs, not a solved problem.
Why it matters — and what’s next
Too heavy, and you risk entrenching incumbents — only the largest can afford compliance — while slowing benefits and pushing development elsewhere. Too light, and harms land on the public, accountability arriving after the damage. Everyone wants the balance point; where it sits is a value judgment. And why is it so hard? Partly because the target moves — capabilities shift faster than laws can be written. Which exposes the assumption underneath: to govern AI, you need a view of what it will become. And on that, serious people disagree profoundly. What are we actually governing? The AGI debate, next.
This is a supplement in AI: Zero → Frontier — a side-trip that deepens the act it sits beside, one file and one loop at a time.
Full transcript 2:45 of narration
Once you accept that AI will reshape work and more, a harder question follows: who governs it, and how? There's no single answer — different societies are trying different philosophies at once. This is the map, laid out neutrally: the major approaches, the trade-off underneath them, and why thoughtful people land in very different places.
Start with the most comprehensive: the EU's AI Act, the first broad AI law of its kind. Its core idea is risk tiers. Most uses are minimal risk, with light obligations.
High-risk uses — hiring, credit, safety-critical systems — face real requirements. A few are banned. It regulates applications by risk, not the technology in the abstract.
The United States has taken, to date, a more sectoral, lighter-touch path — existing agency authorities, executive action that shifts between administrations, and a patchwork of state laws, rather than one federal statute. The instinct: avoid slowing innovation, address harms where they surface. Less unified than the EU's framework, more fluid.
Beneath the specifics sit two philosophies. Precaution: where stakes are high, ask developers to show safety before deployment, accepting slower movement. Permissionless: let innovation proceed and fix harms as they arise, betting openness and speed beat restriction.
Most regimes blend them. The disagreement is about the mix. One tension cuts through all of it — the open-versus-closed question from the open-weights episodes.
Closed models are easier to monitor and restrict, but concentrate capability and are harder to inspect from outside. Open weights bring transparency and competition, but once released can't be recalled, and misuse is harder to prevent. Safety and concentration pull opposite ways.
Which frames the core trade-off. Too heavy, and you risk entrenching incumbents — only the largest can afford compliance — while slowing benefits and pushing development elsewhere. Too light, and harms land on the public, accountability arriving after the damage.
Everyone wants the balance point; where it sits is a value judgment. And why is it so hard? Partly because the target moves — capabilities shift faster than laws can be written.
Which exposes the assumption underneath: to govern AI, you need a view of what it will become. And on that, serious people disagree profoundly. What are we actually governing?
The AGI debate, next.