Artificial intelligence looks certain to revolutionize medicine, but research from Google Cloud suggests it may be more challenging than many people suspect.
Jia Li, who leads research and development at Google Cloud, revealed new research on applying AI to radiology imaging today at EmTech Digital, a conference in San Francisco held by MIT Technology Review.
The work, described in a paper published online, shows that machine learning could help identify disease in a real clinical setting, where data is sparse and where doctors require a rationale for diagnosis. Some AI experts have suggested that whole areas of medical work, such as analyzing radiological images, could become entirely automated thanks to AI.
Li and colleagues used deep learning, a popular machine-learning technique, to identify abnormalities in chest x-rays. Because they had only a small training data set to work with, they used another data set to bootstrap the learning process. They also ensured that their method highlighted the area of an image that was critical to a diagnosing an abnormality. This is important because deep learning is so mathematically complex that it is inherently opaque (see “The dark secret at the heart of AI”).
Li, who is working with medical experts at Stanford, also says that AI can automate only a small part of radiologists’ work. They need to understand a patient’s specific case history, communicate a diagnosis, and determine the right treatment, she says, and they will play a key role in developing accurate and effective machine-learning systems.
So Li believes doctors will not be wholly replaced by AI any time soon. “We can assist [doctors] to make better judgments, and make the process more efficient,” she told the EmTech audience.
The work is important because it highlights and seeks to address key challenges in applying AI to real-world medical situations. Most demonstrations of AI for medicine have involved large, perfectly annotated data sets, and they have not considered the broader context.
The research also suggests that medicine may be a major focus for Google’s cloud platform. Google and others believe that delivering AI through the cloud will be a big, lucrative trend in computing in coming years (see “How the cloud could produce the richest companies ever”).
Researchers like Li also present the technology as a way of “democratizing AI,” or making the technology available to those who don’t have AI training. “Hopefully, experts will spend less time on repetitive tasks,” she said.
A Roomba recorded a woman on the toilet. How did screenshots end up on Facebook?
Robot vacuum companies say your images are safe, but a sprawling global supply chain for data from our devices creates risk.
The viral AI avatar app Lensa undressed me—without my consent
My avatars were cartoonishly pornified, while my male colleagues got to be astronauts, explorers, and inventors.
Roomba testers feel misled after intimate images ended up on Facebook
An MIT Technology Review investigation recently revealed how images of a minor and a tester on the toilet ended up on social media. iRobot said it had consent to collect this kind of data from inside homes—but participants say otherwise.
How to spot AI-generated text
The internet is increasingly awash with text written by AI software. We need new tools to detect it.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.