Will A.I. go Nuclear? ✌️
It's not often A.I., Nuclear and World Peace are used in the same sentence, but I thought it would be interesting to discuss the potential path we may be on toward a better world.
TL;DR
Ever heard of the“Megatons to Megawatts” program? Neither did I until I started researching if we could use uranium from weapons as fuel for reactors. (The answer is yes).
Apparently, the Megatons to Megawatts program was a nuclear non-proliferation program that ran from 1993 to 2013 and is considered a great success. The program converted 500 metric tons of Russian warhead-grade uranium (HEU) into low-enriched uranium (LEU), which is equivalent to eliminating 20,008 nuclear warheads.
What does this have to do with A.I.? Well maybe the exponential increase in demand for energy will bring everyone back to the table to relaunch what seems to be one of the few US/Russia government programs I have read about that was deemed a success by everyone. One could argue we should have never stopped in 2013, and now could be the best time to be thinking this program may help us solve a series of challenges we are facing.
Let's do some math
It takes about 27 tons of Uranium to power a large city per year. (1,000 MW)
The program above (M2M) generated around 15,000 tons of low grade nuclear fuel for reactors. (Which by the way powered 10% of the U.S. electricity for 20 years). So essentially we got 0.75 tons of fuel per warhead.
There are around 15,000 warheads in the world now (mostly Russia and U.S.).
Seems to me, that perhaps this is a discussion that should be on the table. I am no expert, so I would love to get any feedback and comments on what I may be missing - but maybe there is potential check these off our list?
✅ Provide clean energy
✅ Support the requirements to power A.I. which has major potential to advance science, improve access to better education, help cure diseases, unlock solutions for world hunger and grow our production and economy.
✅ Get rid of the 15,000 bombs remaining. Potentially end wars (current and future).
✅ Reduce dependancy on Oil.
Am I crazy to think all we need to do is blow the dust off the old agreement that everyone seemed to like from 1993 and spend the next 20 years on a road toward peace?
The Future Energy Demand of AI and the Role of Nuclear Power
As we edge closer to achieving Artificial General Intelligence (AGI), the demand for computational power is set to surge exponentially. This monumental leap in technology, akin to the discovery of oil or the advent of the internet, is transforming the global landscape, creating new demands for infrastructure and energy. In this blog, we'll explore how the pursuit of AGI is poised to reshape our energy consumption patterns and why nuclear energy may become essential to power the future.
The Compute Revolution: From Bitcoin to AI
The notion of compute as the new oil isn't new; it began with Bitcoin. Bitcoin mining, reliant on Application-Specific Integrated Circuits (ASICs), demonstrated the immense computational power needed to sustain a decentralized digital currency. This phenomenon highlighted the potential and pitfalls of massive compute power—ushering in an era where compute capacity is a critical resource.
Today, we see a similar, albeit far more significant, trend with AI. Major corporations like Google, Facebook, Microsoft, Tesla, Nvidia, OpenAI, and Apple are investing billions in AI infrastructure. These investments are laying the groundwork for AGI, an intelligence that surpasses human capabilities and can perform any intellectual task with efficiency.
The Imminent Demand Surge
While current AI applications like ChatGPT have already pushed the boundaries of computational demand, the integration of AI into everyday tools and devices by tech giants will exponentially increase this demand. Google's Gemini in its apps, Apple's AI in its operating systems, Tesla's autonomous driving AI, and Microsoft's AI-enhanced software illustrate the beginning of this new era.
This impending demand surge draws parallels to historical periods of rapid technological adoption, such as the industrial revolution and the rise of the internet. However, the scale and speed at which AI will permeate every aspect of life will far exceed these past events. The real challenge will be sustaining the energy needs of this compute-intensive future.
Geopolitical and Economic Implications
The comparison of compute to oil brings with it not only economic but also geopolitical implications. As nations strive to gain an edge in AI technology, the demand for computational power could become a strategic asset. Similar to how oil-rich nations have wielded significant global influence, countries with advanced AI capabilities could shape international relations and economic policies.
Furthermore, this dynamic mirrors the nuclear arms race, where the balance of power hinged on technological supremacy. With AI, the stakes are equally high. The European Union's stringent regulations on AI usage aim to ensure safety, but they also risk stifling innovation and driving tech companies to more permissive jurisdictions. This regulatory tension will likely intensify as AI's potential—and its risks—become more apparent.
The Case for Nuclear Power
To sustain the massive energy requirements of AGI, we must look beyond traditional energy sources. Renewable energy, while critical, may not be sufficient to meet the colossal and consistent demand AI will impose. This is where nuclear energy comes into play.
Nuclear power offers a stable, high-output energy source that can provide the continuous power necessary for AI data centers. Unlike fossil fuels, nuclear energy produces minimal carbon emissions, aligning with global sustainability goals. Modern advancements in nuclear technology, such as small modular reactors (SMRs) and fusion research, promise safer and more efficient energy solutions.
Conclusion
The path to AGI is not just a technological journey but an energy one as well. The future of AI will require unprecedented amounts of computational power, reshaping our energy infrastructure and policies. Nuclear energy stands out as a viable and essential solution to meet these demands sustainably.
As we navigate this transformative period, it is crucial to balance innovation with regulation, ensuring that AI's benefits are harnessed while mitigating its risks. The synergy between advanced computational technology and sustainable energy sources like nuclear power will define the next chapter of human progress, paving the way for a future where AGI can thrive.
Hi there, Rich, we met years ago at an entrepreneurship function at New College. I'm glad to see your mind on the cost-side of AI while many selectively tout the benefits. I agree that nuclear will be necessary to support growth in AI while staying on track with emissions goals. I also agree that a program like "Megatons to Megawatts" is perfect in theory, but to reinstate it would be difficult if not impossible with today's international relations. A more achievable goal may be to invest in and better utilize domestic nuclear infrastructure to supply the power for AI, but there's challenges to that too.
In terms of demand, I'd like to point out that, due to gains in processor and cooling efficiency, exponential growth in computing capacity may only drive linear growth in demand for power. Mega-tech companies market the notion that efficiency gains and investments in renewable utilities can keep their emissions stable or even in a downtrend through this process. To pull this off, they need to be able to invest incrementally in power as the demand increases materialize. Since these quantities are hard to predict, something modular like solar is less financially risky than committing to a PPA with a nuclear plant.