China's AI hospitals will transform medicine across the world. But not in the United States.
Summary:
China has invested heavily in AI across medical applications.
Population-scale testing is now underway at Beijing's Tsinghua University, and in Shanghai at Tongji University.
Dramatic improvements are already evident in the observed error rates in diagnosis and treatment recommendations. In the US, diagnostic errors across all disease is over 11%, with high percentages of Americans reporting significant diagnostic errors in their own patient histories. At Tsinghua's AI hospital, errors in diagnosis are under 5%.
As Large Language Models improve and are employed in ever-larger populations of real patients, China's AI in medicine will be rolled out across China, and to dozens of other countries and in areas with limited access to top-quality medical care.
The medical care industry in the United States, however, will almost certainly stand in strong opposition to any tech advances which would reduce the compensation of doctors there. The American Medical Association, along with licensing boards and medical schools, act in concert to strictly ration medical care, and prices and doctor pay as high as possible.
Report:
Artificial Intelligence is attracting trillions of dollars of new capital, and the objective of AI technology is not to automate low-skill tasks. Those have largely already been automated. The application and development of large language models, or LLM’s, and building giant libraries of data for machines to learn from, is to develop applications that can perform high-value tasks. If your company can build robots that can wash dishes, you’ll make a lot of money. But if you build machines that can perform surgeries, you’ll make all the money. The Chinese are building AI hospitals, where the objective is to replace doctors. China is using their AI in medical systems to do more work, do it faster, and do it more accurately.
We reported on it before, when the testing phase for it began, and now they are further along, ready to work with the public. The AI hospital is a partnership between Tsinghua University and a group of companies, and it has 42 AI doctors in 21 departments. They’re ready for public testing in the next 3 months.
China acknowledges that they’re behind the United States in AI spending and research, but experts here in China believe they are going to catch up pretty fast. The factories are here that can more readily deploy AI into real-world physical products, like robots and cars. And it’s just a giant market, with huge populations of data to work with.
So at this hospital they’ve been building the LLM’s for medical systems, and the key procedures of hospital work—registration, consultation, and examination. The doctor-agents are being trained to do the work of actual doctors in diagnosis and treatments.
Each of the AI doctors can work with tens of thousands of patients over several days, which would take years for a human doctor. There’s another project at Shanghai East Hospital and Tongji University, and their LLM—MedGo—was fed over 6,000 medical textbooks and enormous data sets of patient data.
So far, the AI doctors at Tsinghua have achieved high degrees of accuracy: 88% accuracy in examination, 95% in diagnosis, and 77% in treatment. And those are high levels. Tsinghua’s error rates in diagnosis would be 5%. The Harvard Gazette found that among ICU patients in the United States, 23% experienced diagnostic error, the majority which left the patient worse off. Johns Hopkins and BMJ published a study that shows about 800,000 Americans die or are disabled by diagnostic errors every year. The overage error rate across all diseases was 11.1%, with spinal disorders showing error rates of 62%. For stroke, 17.5% of patients had errors in diagnosis. A survey of patients revealed that 55% of them claimed that errors in diagnosis was their biggest concern about seeing a doctor, and a survey of doctors shows that half of doctors see diagnostic mistakes at least once a month.
And this is another reason that the tech investors in China are so optimistic about the medical applications for their AI. Just from what we know, right now—where do you think you’ll be better off, as a patient, getting a proper diagnosis and treatment recommendation? A hospital with a human doctor, who may be tired, or going through a divorce, or in a hurry, or hungry, or maybe not current in his field? Or one of these AI doctor-agents, who has memorized 6,000 medical school textbooks, along with all the new research that comes along in the future, and who can scour real-life medical records of a billion people who live in China, to try to figure out what your problem probably is?
There is another issue with respect to market friendliness. To foreigners it may sound a little off, when Chinese officials claim to have advantages in this area. But consider just this tech, here, for hospitals and medical staff replaced by artificial intelligence. It will be a transformative, and disruptive, technology, rolled out in scale first across China, then in the developing world where there are shortages of top doctors. China knows they’ve got a market for it not just here, but in most of the world, once they get it figured out.
But in the United States, we should expect a lot of resistance from the American Medical Association, and medical schools, and especially from doctors themselves. Doctors in the United States make many times what doctors do almost anywhere else, and so the entire system is highly motivated in making sure that this technology is NOT rolled out there. Ironically then it’s much more likely that 10 years from now, patients in the developing world will receive more accurate diagnoses, better treatment options, more personalized care, than a patient in one of the top hospitals in the US, who is paying a lot more besides.
Resources and Links:
South China Morning Post, Tsinghua University-incubated start-up to widen test of virtual hospital with ‘AI doctors’
https://www.scmp.com/tech/tech-trends/article/3289015/tsinghua-university-incubated-start-widen-test-virtual-hospital-ai-doctors
SCMP, China’s AI integration ‘advancing rapidly’ as it plays catch-up with US https://www.scmp.com/economy/economic-indicators/article/3277783/chinas-ai-integration-advancing-rapidly-it-plays-catch-us
SCMP, China’s AI industry could see US$1.4 trillion in investment in 6 years, executive says https://www.scmp.com/tech/big-tech/article/3277743/chinas-ai-industry-could-see-us14-trillion-investment-6-years-executive-says
Harvard Gazette, High rate of diagnostic error found in ICU https://news.harvard.edu/gazette/story/2024/01/research-assesses-rates-causes-of-diagnostic-errors/
Johns Hopkins, Report Highlights Public Health Impact of Serious Harms From Diagnostic Error in U.S.
https://www.hopkinsmedicine.org/news/newsroom/news-releases/2023/07/report-highlights-public-health-impact-of-serious-harms-from-diagnostic-error-in-us
NIH, The incidence of diagnostic error in medicine https://pmc.ncbi.nlm.nih.gov/articles/PMC3786666/
Will the $1 trillion of generative AI investment pay off?
Goldman Sachs, Will the $1 trillion of generative AI investment pay off? https://www.goldmansachs.com/insights/articles/will-the-1-trillion-of-generative-ai-investment-pay-off
Bloomberg, Generative AI to Become a $1.3 Trillion Market by 2032, Research Finds https://www.bloomberg.com/company/press/generative-ai-to-become-a-1-3-trillion-market-by-2032-research-finds/
YouTube, Inside China Business, China, and not the US, just built the world's first Artificial Intelligence hospital. Why, and why?
The Medical Cartel is Keeping Health Care Costs High
https://fee.org/articles/the-medical-cartel-is-keeping-health-care-costs-high/
Thumbnail image, Medical Error Statistics
https://etactics.com/blog/medical-error-statistics