Stay up up to now with free updates
Simply register for Artificial intelligence myFT Digest – delivered straight to your inbox.
Machines are already superior to humans in relation to playing chess, recognizing birdsong, or predicting complex protein structures. But in relation to really clever and intuitive things, like original scientific research, we humans prefer to think we still have the sting.
Maybe we want to re-evaluate. At the RAAIS Artificial Intelligence conference in London earlier this month Daniel Cohen, The president of Canadian drug discovery company Valence Labs spoke concerning the tantalizing, if somewhat disturbing, possibility of “autonomous scientific discovery.” Sophisticated AI models trained on specialized data may soon give you the chance to develop hypotheses, design and run experiments, learn from the outcomes, and repeat this 24/7. “Our mission is to industrialize scientific discovery,” he said.
You don't should discuss with people in computational biology for long to know their enthusiasm for AI. AI research firm Google DeepMind even spun off a separate company, Isomorphic Labs, to reap the benefits of this area after its AlphaFold program modeled 200 million protein structures.
The promise is that computational biology can advance scientific research, speed up drug discovery, and improve treatment outcomes. Machines have plenty of benefits over their flesh-and-blood counterparts, researchers and lab assistants. For one thing, they don't must sleep, or take care of colds, hangovers, or difficult relationships.
“I'm very encouraged by the speed at which the sector is advancing,” Christina Curtis, a professor of genetics and biomedical data science at Stanford University School of Medicine, tells me. “It's changing our understanding of disease, how we detect malignancy, and the way we treat and stop it.”
Curtis was the lead writer of an article, published last month in Sciencethat studies the heritability of malignancy in various kinds of cancer. Using machine learning techniques, the researchers analyzed hundreds of genomes from individuals with preinvasive and invasive breast tumors to look at differences of their immunological response to the disease. They found that the way in which tumor cells developed in individuals was “shaped” by the germline genome they inherited at conception.
Such research may lead to earlier diagnosis and more personalized treatment, improving survival rates. “More than 50 percent of cancer diagnoses are stage 4 or higher. We get the knowledge too late to make decisions,” says Curtis. “Ideally, we will do that more preventively.”
There are two major limitations. The first is that “genetics provides clues, but not answers,” said one industry executive. Machines have identified many targets for drug development, but few successful products have been delivered to market. Even when technology results in scientific breakthroughs, it takes a few years for brand new drugs to realize approval.
Thore Graepel, the worldwide head of computer science at Altos Labs, previously helped develop the AlphaGo program at Google DeepMind. AlphaGo's victory over the world's strongest player at the traditional game of Go was seen as a surprising breakthrough in machine intelligence. But Graepel told the RAAIS conference that the biological complexities he now faces in cell rejuvenation are “orders of magnitude” greater. “I've never seen a lot complexity with so little data,” he said.
The second limitation is data scarcity. Curtis argues that patient data is like “liquid gold” for researchers, but we don't yet have the mechanisms to gather it routinely. The most useful thing can be to mix a patient's genetic information with longitudinal health data collected throughout their treatments and life.
Reorienting health systems towards early monitoring and prevention and away from late diagnosis and treatment would require a profound transformation of cumbersome organisations. But Britain's Labour Party, which is predicted to win next week's general election, is pledging to speed up this transformation of the National Health Service. The Labour Party Manifesto guarantees to establish a “Fit for the Future” fund to double the variety of CT and MRI scanners for detecting early-stage cancer.
Voters are rightly skeptical when politicians make big guarantees. But given the strains on public funds in ageing societies, governments may soon haven’t any selection but to go down this path. As the Dutch philosopher Desiderius Erasmus supposedly told us five centuries ago, “an oz of prevention is best than a pound of cure.” In this respect, AI may very well be certainly one of our best assets.
