The UK's new £225 million National Artificial Intelligence Supercomputer will help create wearable AI assistants to find sick dairy cows in Somerset, improve detection of brown skin skin cancer, and help riot police predict danger.
Scientists hope that Isan Bird-AI, named after Isan Bird Kingdom Brunel, a 19th-century engineer of groundbreaking bridges and railways, unleashes a wave of AI-powered technology, healthcare and society breakthroughs by allowing scholars and public bodies to access the vast computing power of the kinds of guardians of private technology companies.
The supercomputer was officially turned on in Bristol on Thursday by Secretary of Science and Technology Peter Kyle.
The machine is equipped with 5,400 Nvidia “Superchips” and is located in a black metal cage topped with razor wires topped with. It spends around 1 million pounds a month on most nuclear power and drives 100,000 times faster than the average laptop.
In the midst of fierce international competition for computing power, it is the UK's largest publicly recognized facility, but it is the 11th fastest after the world of the US, Japan, Germany, Italy, Finland and Switzerland. Though Elon Musk's new Xai supercomputer in Tennessee already has 20 times its processing power, Meta CEO Mark Zuckerberg is planning a data center that “covers a significant portion of Manhattan's footprint.”
The investment is part of a £2 billion push to achieve government “AI sovereignty,” so the UK does not need to rely on foreign processing chips to make advances in AI-enabled research. However, switching on could raise a new ethical dilemmas about how far AI should guide policies, from controlling public protests to breeding animals.
One of the AI models under development by academics at the University of Bristol is an algorithm that learns from thousands of hours of footage of human movements captured using wearable cameras. The idea is to try to predict how humans can move next. It can be applied to a wide range of scenarios, such as allowing police to predict how a crowd of protesters will behave, or to predict accidents in industrial settings such as construction sites.
Dima Damen, a professor of computer vision at the university, said that the algorithm trained by Isambad AA, based on the patterns that wearable cameras are capturing in real time, based on patterns of human behavior, “can even give an early warning that something is likely to happen in the next two minutes.”
Damen added that there is “the big ethical implications of AI” and that it is important to always know why the system made the decision.
“One of the fears of AI is that some people own technology and know-how, and others don't,” she said. “It is our greatest duty as a researcher to make sure that data and knowledge are available to everyone.”
Another AI model under development may detect early infections in cattle. Somerset herds are being photographed 24 hours a day, and models are trained to predict whether animals are at an early stage of mastitis, an animal welfare issue. Bristol scientists believe this is possible based on detecting subtle changes in cows' social behavior.
“The farmers clearly have a great interest in the herd, but they don't have the time to continue to see all the cattle in the herd during the day, so AI is there to offer that view.”
A third group of researchers use supercomputers to detect bias in the detection of skin cancer. James Pope, a senior lecturer in data science at the University of Bristol, runs “billions, if not a quarter-dollar calculation” in Isamubird, finds that the current phone app checks for malls and lesions of cancer signs. If confirmed in further testing, you can re-adjust the app to avoid bias.
“It's extremely difficult and frankly it's impossible to do it on a traditional computer,” he said.
