The World and AI

Essay 7 — Understanding AI Course

So far in this course, we've looked at what AI is, how it works, and how individuals and organisations can use it. But AI isn't just a personal tool or a business opportunity — it's a global phenomenon with implications that stretch from geopolitics to the environment. In this final thematic essay, we'll zoom out and look at the big picture: the international politics of AI, its ecological footprint, the dominance of big corporations, and the growing divide between those who have access to AI and those who don't.

AI as a Geopolitical Battleground

AI has become a major arena of international competition — and in some ways, a new Cold War.

The two dominant players are the United States and China. Both have invested enormous resources into AI development, and both see AI leadership as central to their future power and influence.

The US perspective

The US currently leads in the most advanced AI research, largely through companies like OpenAI, Google DeepMind, Anthropic, and Meta. The US government has taken an interest too — not just funding research but also restricting China's access to advanced AI chips and the equipment needed to make them.

China's perspective

China has stated explicitly that it wants to be the world's leading AI power by 2030. It has invested heavily in AI research, deployed AI extensively in governance and surveillance, and has its own powerful AI companies (Baidu, Alibaba, ByteDance, and others). China's approach tends to be more state-directed than the US model.

The implications for the rest of the world

For the UK and Europe, this competition creates pressures and choices:

• Which side do you align with? Neither — but navigating between US and Chinese AI ecosystems is becoming more complex
• Strategic autonomy: Can Europe develop AI capabilities independent of both superpowers?
• Standards: Who sets the global standards for AI — technical, ethical, and regulatory?
• Dependencies: Many countries rely on US AI infrastructure. What does that mean for sovereignty?

The EU has attempted to chart its own path with the EU AI Act — the world's first comprehensive AI law. The UK, post-Brexit, has taken a more sector-specific, light-touch approach — trying to position the UK as a good place to develop AI, rather than regulate it heavily.

The Ecological Cost — AI's Hidden Footprint

Here's something that rarely makes the headlines: AI is enormously energy-hungry.

Training a large AI model requires huge amounts of computing power — and that computing power uses electricity. A 2023 study estimated that training GPT-3 consumed roughly 1,287 megawatt-hours of electricity — equivalent to what a US household would use in 130 years. More advanced models are significantly more power-hungry.

The carbon footprint

The data centres that run AI services are among the most energy-intensive facilities on Earth. Data centres in Ireland, for example, now consume around 20% of the country's electricity — and AI demand is growing fast.

Water, too

Less discussed, but equally important: data centres also use enormous amounts of water for cooling. Microsoft's water consumption increased by 34% between 2021 and 2022 — largely driven by AI data centre cooling needs.

What can be done?

• More efficient models: Researchers are working on smaller, more efficient models that perform almost as well as larger ones at a fraction of the energy cost (this is sometimes called "efficient AI" or "small language models")
• Renewable energy: Major tech companies have committed to matching their data centre energy use with renewable sources — though progress is mixed
• Local AI: Running smaller models locally (Essay 3) can have a significantly lower environmental footprint than using cloud-based AI
• Thoughtful use: Not every task needs the most powerful AI. Using a smaller model for simple tasks reduces unnecessary energy use

The Corporate Dominance Problem

One of the most significant structural features of the current AI landscape is how concentrated power is in a small number of very large technology corporations.

Who controls AI right now?

A handful of American companies — Alphabet (Google), Microsoft/OpenAI, Meta, Apple, Amazon, and a few others — largely control the most advanced AI research and deployment. In China, it's similarly concentrated: Baidu, Alibaba, ByteDance, and Tencent.

This concentration raises serious questions:

Accountability: When a small number of corporations control transformative technology, who holds them accountable? Governments are struggling to keep up. Users have very little leverage.

Incentives: These companies have shareholders to satisfy and profit to make. Their AI is designed partly to serve users — but also to generate revenue. Those interests don't always align.

Competition: It's extremely hard for smaller companies or academic institutions to compete with firms that spend billions on AI research. Some experts worry this will lead to a static, monopolistic tech sector.

Influence: Large AI companies have enormous influence over what the technology looks like, what it can do, and what it cannot do. They set the defaults, the content policies, the capabilities, and the limitations. Most of the world uses AI on their terms, not ours.

Is this inevitable?

Not entirely. The open-source AI movement — particularly Meta's decision to release its Llama models — has significantly disrupted this concentration. Local AI (Essay 3) represents another dimension of resistance to corporate control. And governments have some power to regulate and redistribute through competition law and AI-specific regulation.

But the structural advantages of the major players are enormous, and the current trajectory points toward continued concentration of AI power in a small number of corporations.

Access to AI — The Digital Divide, Again

The digital divide — the gap between those who have access to technology and those who don't — has been a concern since the internet became mainstream. AI is creating a new layer of this divide.

The geography of AI

AI access is deeply unequal across the world:

• High access: US, UK, Western Europe, China, South Korea, Japan — where the major AI companies are based and the infrastructure exists to use AI extensively
• Limited access: Much of Africa, South Asia, and parts of Latin America — where computing infrastructure is limited, electricity is unreliable, and broadband access is scarce
• No meaningful access: Some of the world's poorest regions, where even basic internet access is unavailable

Within countries, too

Even in wealthy countries, AI access is unequal. Consider:

• A university student with a laptop and a Copilot subscription has enormous AI-powered research and writing capability
• A care worker on minimum wage, without a personal computer, has virtually none
• A well-funded school can give its students AI tools as part of education
• A poorly funded school cannot

This isn't just about the technology — it's about who benefits from it. If AI makes knowledge workers dramatically more productive, and that productivity gain flows disproportionately to those who already had advantages, inequality grows.

The concentration of data

There's another layer: AI systems are trained on data. And the data used to train them reflects the world as it is — largely generated by people in wealthy, English-speaking countries. This means AI often works better for people like that — and less well for people who are different.

What Can Be Done?

These are big problems. But they are not unsolvable. Here's what various actors can do:

Governments can:

• Regulate AI thoughtfully — not just for safety and rights protection, but to ensure competitive markets and prevent monopolistic concentration
• Invest in public AI infrastructure — some researchers and politicians have proposed national or regional AI computing facilities that are publicly owned and accessible (the UK has begun discussing this)
• Support digital inclusion — ensuring schools, libraries, and community centres have AI access
• Require transparency from AI companies about what their systems do, what data they use, and what their limitations are

International bodies can:

• Develop global norms and standards — the UN, OECD, and G7 have all begun working on AI governance frameworks
• Bridge the access gap — initiatives to bring AI infrastructure and education to lower-income countries
• Monitor corporate power — through competition law and international agreements

Companies can:

• Open their models — Meta's Llama approach shows this is possible
• Invest in efficiency — smaller, more efficient models reduce the environmental footprint and can be deployed more widely
• Be transparent — about training data, limitations, and biases

Individuals can:

• Use local AI where possible — reducing dependence on corporate infrastructure
• Stay informed — understanding what's happening with AI is itself a form of power
• Advocate — for AI policies that serve the many, not just the few

What Have We Learned?

• AI has become a major arena of geopolitical competition between the US and China, with significant implications for the rest of the world
• AI has a substantial and growing environmental footprint — data centres consume enormous electricity and water
• A small number of very large corporations largely control the most powerful AI systems, raising serious accountability and competition questions
• Access to AI is deeply unequal — both between countries and within them, mirroring and potentially worsening existing inequalities
• Governments, companies, and individuals all have roles to play in shaping AI's future

Glossary of Terms

Further Thinking

• Should AI be treated more like a public utility (like electricity or water) and regulated accordingly? Why or why not?
• Who should set the rules for AI — governments, the companies that build it, or some combination?
• What would a world where AI access was truly equitable look like? How far are we from it?

Essay 7 of 8 — Understanding AI Course
Authors: Bea Groves-McDaniel and SAL-9000
Licensed under Creative Commons (NC, ND) — Share with attribution, no commercial use.