In a recent episode of the 80,000 Hours podcast, Will MacAskill talked about an economic future where advanced AI systems could trigger an intelligence explosion – a period of unprecedented technological development that could compress a century of progress into just a decade.

After listening there were still some open questions to me about the constraints of natural resources and the physical limits of growth. I've tried to answer these questions myself in the post below based on the podcast interview.

Please feel free to comment. I'd very much like to have some input about these thoughts.

Types of intelligence explosion

Will MacAskill distinguishes between 3 types of intelligence explosions that could occur as AI advances:

1. Software intelligence explosion: AI systems become increasingly capable of designing better AI algorithms, creating a feedback loop of improvement.
2. Technological intelligence explosion: AI excels at designing better chips and hardware, dramatically increasing computational power.
3. Industrial intelligence explosion: AI and robotics enable fully autonomous factories that can produce more factories, creating exponential growth in manufacturing capacity.

He believes all three are important, although the industrial explosion may be the most consequential for geopolitics and economic power.

Manufacturing advantage

Countries with existing manufacturing infrastructure (e.g. China) could gain many advantages in this new world because they could rapidly building out industry:

1. Authoritarian countries can have higher savings. They could invest those outputs again as inputs by building even more factories (high reinvestment rates prioritized industrial growth over consumption).
2. Countries that remain heavily industrialized rather than transitioning to service economies would have a head start, because they will be able to scale up the manufacturing very fast.

If a country gains a lead in autonomous manufacturing capabilities, that lead could compound at an accelerating rate, potentially creating an unstoppable advantage.

Growth of autonomous manufacturing

What makes this different from the previous industrial revolution is the super-exponential growth curve because of the autonomous industry (AI and robots doing everything).

"How many goods do you have if you have two factories? It seems like twice as many. Standard economic view."

If some of those goods include improvements to the factories themselves or more factories, you get an effect beyond traditional growth models. If manufacturing systems could replicate at similar rates as fruit flies – doubling every week – growth would become practically unimaginable by today's standards.

Resources?

But this raises the resources question:

Wouldn't physical resource constraints severely limit this growth?

Even the most efficient factories cannot manufacture goods out of nothing. They require steel, aluminum, silicon, rare earth minerals, plastics (petrochemicals), and a lot of energy.

Will MacAskill acknowledges these constraints indirectly when discussing limits to growth. He sees Earth-bound resources as a meaningful but temporary constraint – one that could eventually be surpassed through space-based resource utilization. And he does mention environmental constraints by warning for a thermodynamics heat-up of the planet because of the energy use of the race to build up greater industry.

But this still doesn't solve the resource question.

Why resource constraints might not apply

1. Efficiency revolution

Current manufacturing is nowhere near theoretical efficiency limits. Advanced AI could dramatically reduce waste in resource extraction, processing, and manufacturing. Materials that are currently discarded due to imperfections or inefficient processes could be utilized.

For instance the buy-to-fly ratio for aerospace components (raw material to finished product) can be 20:1, meaning 95% of material is wasted. AI-optimized processes could potentially approach 1:1 ratios.

2. Material substitution

AI could accelerate materials science, enabling rapid discovery of alternatives to scarce resources, like substitutes for rare earth minerals in electronics, as happened earlier (synthetic rubber, aluminum replacing copper...).

3. Circular economy at scale

Current recycling systems capture only a fraction of the potential reusable materials in our waste streams. AI-managed systems could approach 100% recovery rates for many materials, effectively creating closed-loop systems.

4. Extraction improvements

We currently extract resources from a tiny fraction of the Earth's crust and oceans. AI could enable economical extraction from currently unreachable sources (deeper mines, lower-grade ores, ocean floor mining, extraction from seawater...).

5. Energy potential

Energy is perhaps the most fundamental constraint on industrial activity. Technologies as fusion, improved solar, other novel energy sources could dramatically increase available energy. 

6. Space resources

Will MacAskill mentions also space as an unlimited source of resources. Near-Earth asteroids alone contain quintillions of dollars worth of metals.

Timescale?

The most important factor may be the timescale of the intelligence explosion versus the timescale of resource depletion. If AI accelerates technological development fast enough, new solutions to resource constraints might emerge before those constraints become binding.

Question: will resource constraints slow growth before technological advances remove those constraints?

Conclusion: China in the lead

Advanced AI could indeed revolutionize manufacturing and potentially create unprecedented growth rates, but physical resource constraints would almost certainly play a significant role in timing.

The question is maybe not whether resource constraints exist (they do), but if technological acceleration could overcome them faster than they become binding. This remains an open question.

What seems most likely is that countries with both strong manufacturing bases ánd secure access to critical resources will have the greatest advantages. This suggests that China's strategy of developing manufacturing capabilities while securing resource supplies globally (Africa!) may be even more proactive than I had realized.