AS a former physician, I shivered a bit when I heard Dr. Vivek Wadhwa say he would rather have an artificial-intelligence doctor than a human one. “I would trust an A.I. over a doctor any day,” he proclaimed at a recent health innovation conference in San Francisco, noting that artificial intelligence provided “perfect knowledge.” When asked to vote, probably a third of those in attendance agreed.

But it made sense: Dr. Wadhwa is a professor, entrepreneur and technology visionary. What’s more, the conference took place in San Francisco, where faith in the power of technology and data to solve problems holds unshakable sway.

There was certainly plenty of innovation on display at the conference’s rooftop reception, called “Health by the Numbers”: One device attaches to your iPhone and turns it into an otoscope so you can see if your child has an ear infection; another allows it to check your blood alcohol level. Attendees could check out homecholesterol test kits, and a wearable device to track the “quality” of their breathing.

Silicon Valley is bringing a host of new data-driven technologies to health care, many of them with enormous potential. But before we rush to measure every human attribute in real time, it would be a good idea to ask: When is more data actually useful to promote and ensure better health? And when does technology add true value to health care? The results have been mixed.

“It holds great promise and excitement, but so far everyone is often disappointed with the outcomes,” said Steven J. Van Kuiken, a director at McKinsey and Company who studies technology and health care. “There are lots of interesting ideas, but how do you get data that’s useful to patients, physicians or regulators? And then how do you get them to act on it?”

While the proliferation of fitness trackers suggests there is commercial potential in consumer health data technology, utility may be limited. “I don’t doubt the wearable piece is going to be a productive business model for people,” Ian T. Clark, chief executive of Genentech, said at the conference. “I just don’t know whether it’s going to bend the curve in health outcomes.”

Last month Aetna announced it was discontinuing CarePass, its personalized health data platform for patients, because it hadn’t delivered on anticipated results. And some studies show that half the people who buy portable fitness trackers stop using them in a matter of months. That is probably because most people who wear them are already health-conscious and there may be little long-term value once they take note of their activity patterns, Mr. Van Kuiken suggested.

So how can we create innovative new technologies that will revolutionize health rather than end up as discarded Christmas presents?

One big challenge is that the elusive state we call “health” is not always easily measurable. Normal blood pressure jumps up and down in response to thoughts, hydration and stress. Some healthy people have low platelets or slight elevations of liver enzymes.

In some cases, the ability to collect data has outpaced medical understanding. There is no “normal” testosterone level for an aging male, yet millions of men have been told they suffer from a condition called Low T and are using testosterone gels, even though medical studies have shown the products are dangerous.

In other cases, the data can show up fine, even when the patient is not. You don’t always measure the right thing. When I was in medical school, there was a gallows joke that some patients die with “Harvard numbers.” In other words, the lab tests that were ordered were all perfect, but the patient died anyway.

On the other hand, the test results can look bad even when the patient is fine. Scans of the spine, for example, show that many people have big bulging discs but no back pain. So which do you treat, the M.R.I. or the patient?

Can a Computer Replace Your Doctor?