THE FIELD of AI has changed dramatically in recent years, with an explosion of systems like those behind the blockbuster hit ChatGPT creating a flurry of investment and dozens of new start-ups. But the power of AI isn鈥檛 only being wielded to make chatbots. In fact, it has been quietly helping us solve many of our biggest problems for decades, and now even looks set to provide fresh impetus in the quest to achieve commercially viable nuclear fusion power.
One company that has consistently shown a knack for solving real-world problems is Google鈥檚 DeepMind. One of its most surprising achievements has been in protein folding. Determining the crumpled shapes of proteins based on their sequences of constituent amino acids had been a persistent problem for decades, with researchers often taking years to solve a single one. DeepMind transformed biology last year by announcing that it had predicted the structure of nearly all proteins known to science in just 18 months. The team had trained its AI AlphaFold on data from known protein shapes and it learned to predict what the unsolved proteins would look like.
Advertisement
Future medicines
The data is already helping researchers make advances in everything from to the creation of . at DeepMind says there is even more on the horizon. 鈥淏eyond protein structure prediction, there is more work to be done mapping protein dynamics, accelerating protein design and understanding the effect of protein mutations 鈥� for example, those associated with diseases like cancer,鈥� he says.
AI is also making vital inroads into the problems associated with drug development. That process involves collecting and analysing disparate types of data from lab experiments, computer simulations, scans, clinical trials and health records. These days, it takes ever longer and costs ever more to bring new drugs to patients, perhaps because the low-hanging fruit has already been picked and regulation has increased. But several projects are now using AI to automate parts of the process, says Conrad Bessant at Queen Mary University of London, such as taking large, messy datasets and organising them in a way that makes analysis easier 鈥� or using AI to write code to do the job. Researchers are also using generative AI to produce molecular structures that they think would be useful in targeting certain conditions.
Elsewhere, our efforts to combat one of the world鈥檚 biggest problems 鈥� climate change 鈥� are also getting a helping hand. Here, AI is being used to create more energy-efficient cars, computers and even wind turbines. For instance, recent work from researchers at the Polytechnic Institute of Paris used AI to ensure that turbines were pointed into the wind more often, boosting output by 0.3 per cent. This may seem small, but if it was rolled out globally it could be enough to power the equivalent of 1.7 million UK homes.
DeepMind has also developed AI to improve standard computing tasks like matrix multiplication, which it boosted by up to 20 per cent, and sorting algorithms, which it speeded up by as much as 70 per cent. Both tasks are performed trillions of times a day on computers around the world. Together, these seemingly small steps provide significant contributions to helping us edge towards net zero by reducing emissions from computing.
Facebook owner Meta, meanwhile, has used AI to develop a process for creating concrete that emits 40 per cent less carbon. As concrete is responsible for up to 8 per cent of global carbon emissions, this could be a huge help in our fight against climate change. And this is just the latest in a long line of counterintuitive engineering designs developed by AI that have outperformed human endeavours. Back in 2006, for example, that met all the necessary criteria for the mission.
Finally, we come to fusion. Researchers have been trying to create an efficient, reliable nuclear fusion power plant for decades, with the promise that a breakthrough will make energy so cheap that it could be given away for free. But it is extremely challenging. So much so that there is a saying: fusion power is just 20 years away, and it always will be.
Inside a tokamak fusion reactor, multiple magnetic coils work to keep the plasma 鈥� which is hotter than any part of the sun 鈥� safely contained. Precisely and rapidly controlling multiple coils to compress the plasma into a confined shape, safe from catastrophic contact with the walls of the machine, is maddeningly difficult.
Now, AI hasn鈥檛 actually solved the problem, but it is helping. Last year, researchers at DeepMind and the Swiss Federal Institute of Technology in Switzerland created a neural network capable of controlling 19 magnetic coils. The AI was also able to sculpt the plasma in a tokamak into various shapes at will. at the University of Manchester, UK, says that fusion reactors are now a proved concept. AI could be the turning point to finally make them a reality.
This story is part of a series in which we explore the most pressing questions about artificial intelligence. Read the other articles below
How does ChatGPT work and do AI-powered chatbots 鈥渢hink鈥� like us? | What generative AI really means for the economy, jobs and education 触听Forget human extinction 鈥� these are the real risks posed by AI today 触听Can AI ever become conscious and how would we know if that happens? 触听How to use AI to make your life simpler, cheaper and more productive
听
