We had two covers this week, on the race to push artificial intelligence beyond today’s limits and on how the world’s poorest people are suffering as development stalls.

Two years after ChatGPT took the world by storm, generative artificial intelligence seems to have hit a roadblock. The energy costs of building and using bigger models are spiralling, and breakthroughs are getting harder. Our cover is about how researchers and entrepreneurs are racing to come up with solutions.

AI is a work in progress—and progress is getting harder. By one estimate, today’s biggest models cost $100m to train; the next generation could cost $1bn, and the following one $10bn. Likewise, asking a model to answer a query comes at a computational cost of anything from $2,400 to $223,000 to summarise the financial reports of the world’s 58,000 public companies. It is hard to see how that adds up to a viable business model. 

We thought about illustrating this with a half-finished automaton. But, as so often with AI covers, we immediately ran into the objection that artificial intelligence and robotics are different things. In addition, some of us were reminded of Auguste Rodin’s “The Thinker”—and not in a good way. The sculpture has become something of a visual cliché. 

The backing for this story includes a Technology Quarterly (TQ) which explains how companies are developing chips to run large language models. These are more efficient than more general-purpose processors, such as Nvidia’s, and therefore use less energy and are cheaper.

The cover of the TQ uses an iridescent palette to illustrate a convoy of chips marching towards the mountains abutting Silicon Valley. How about that as a starting-point for the cover? The hand balancing a chip is beautiful. It also has the virtue of being disembodied—virtual rather than android. 

Unfortunately, this cover could be saying almost anything. We had our visual language, but not yet a message. 

AI needs a breakthrough, and here we illustrate that with the hand pushing through a barrier. In a near-finished version we tried the hand coming up from the bottom of the page, but it looks too much as if it is lifting the flap of a tent. 

AI’s problems are not a reason to panic. Plenty of other technologies have faced constraints and gone on to prosper. Getting people into space led to new products that influenced what happened on Earth, too; the oil-price shock in the 1970s spurred on efficiency across the economy. 

Until now, it had seemed as if dominating AI was about mobilising capital, energy and computing power. Now it looks as if the winners will also need talent and a flourishing ecosystem for invention. In what still promises to be a transformative technology, a magic ingredient will determine which firms, investors and countries prevail. That ingredient will be innovation.

In the two decades after around 1995, extreme poverty plummeted, the gaps in GDP between rich and poor countries narrowed, and global public health and education improved vastly. Since 2015 that miracle has ground to a halt. Depending on where you draw the line between rich and poor countries, the worst-off have stopped growing faster than richer ones, or are even falling further behind. For the more than 700m people who are still in extreme poverty—and the 3bn who are merely poor—this is grim news.

This watering can has patches of rust shaped like a map of the world, with water leaking from the global south. And the wheelbarrow is being pushed uphill by a boy struggling to make much progress. Both are fine images, but they lead you to think about aid. 

It is true that in the poorest countries education and health have depended on donors writing big cheques. It is also true that aid budgets have been squeezed, and cash has been diverted from helping the poorest to other causes, such as greening power grids and helping refugees. Of what aid money remains, much is wasted rather than being spent after careful study of what works. 

The thing is, however, that even when aid has improved lives, it has not unleashed sustainable growth.

We preferred the road that suddenly stops, as a symbol of arrested progress. 

In the 1990s global convergence was powered by a few big successes: China’s rapid growth after it opened up under Deng Xiaoping, a similar process in India after reforms dismantling the “licence Raj”, and the integration of countries in eastern Europe into the global market economy after the fall of communism. 

Unfortunately, home-grown reform has ground to a halt. These days, many leaders are more interested in state control, industrial policy and protectionism. It is surely no accident that such policies also strengthen their own grip on power.
 

The rebar bar chart is better, partly because it has a person in it. This sketch needed some work. The last pillar needs to show that growth has stalled; the shadows are wrong; the texture of the concrete is uneven. We fixed those and picked words whose outline pleasingly echoed the height of the bars.

In the era of catch-up, the West preached the virtues of free markets and free trade. As the rich world has turned towards intervention, so the chosen instrument for poor countries has likewise become trade protection and industrial policies. Rich countries will cope with the costs of their own mistakes, as they usually do. For the poorest people, however, growth can be the difference between a good life and penury.