Venture capitalist Vinod Khosla thinks that the best way to improve health care is to get rid of most doctors and replace them with data crunching machines. The point, he argues, is that human judgment cannot compete against machine-learning systems that derive predictions from millions of data points.
“Biological research will be important, but it feels like data science will do more for medicine than all the biological sciences combined,” he said on Friday, during the final day of Stanford University School of Medicine’s Big Data in Biomedicine Conference. “I may be wrong on the specifics, but I think I will be directionally right.”
“Human judgment cannot compete against machine-learning systems that derive predictions from millions of data points.”Big data in health care has the potential to bring big changes to the industry, relying on massive volumes of digital information generated by electronic health records, genetic sequencing, clinical trials and other sources to deliver evidence-backed advice.
“Sufficient data used properly and reduced to the right insights does in fact make up for errors,” Khosla said. “I would rather have 1,500 EKGs done much more poorly than two EKGs done a year very well, because the sources of errors in the current system are just too large. When I have two EKGs a year, I may not be symptomatic. I’m not arguing that these systems don’t have errors. I’m saying the volume of the data, properly applied, makes up for it.”
He went on to add that it’s only a matter of time before health care accepts that technology can do a better job of predicting patients’ risks for diseases, diagnosing illnesses and pinpointing the most effective therapies. In particular, wearable medical sensors like Fitbit can already give patients advance on how to make informed health and health-related decisions on their own.
“It’s only a matter of time before health care accepts that technology can do a better job of predicting patients’ risks for diseases, diagnosing illnesses and pinpointing the most effective therapies.”Not surprisingly, physicians in the crowd didn’t like what they’ve heard, with one of them saying: “I don’t agree with 80 percent of your remarks.”
Khosla acknowledged his view is often not a popular one, but did not back down. “Humans are not good when 500 variables affect a disease. We can handle three to five to seven, maybe,” he said. “We are guided too much by opinions, not by statistical science,” he concluded.