Why we’re teaching computers to help treat cancer

May 12, 2015

PBS Newshour | Every day, we depend on artificial intelligence to help us make sense of a steady deluge of information. AI helps the post office to sort its mail, Wall Street to make financial decisions and physicians to diagnose patients. Hari Sreenivasan reports on how tech firms are investing in the next generation of intelligent computer programs and in what ways the technology still lags behind humans.

related reading:
PBS | Newshour
PBS | YouTube channel: main
PBS | YouTube channel: Newshour

more on this theme on Kurzweil Network:
media — PBS Newshour | How smart is today’s artificial intelligence?

full transcript:

GWEN IFILL: Now we continue our series about artificial intelligence, called AI, where computers are able to make intelligent decisions without human input.

As computing power gets stronger and people continue to generate massive amounts of data, AI is making its way into the marketplace and into your doctor’s examination room.

Hari Sreenivasan has the latest in our series on breakthroughs in invention and innovation.

HARI SREENIVASAN: Advances in artificial intelligence continue to push the boundaries between science fiction and reality, like this brain controlled device at the University of Minnesota. It enables users to fly a model helicopter with only their thoughts. The hope is it will soon help disabled people to operate robotic arms.

But you don’t need to be in a university lab to find AI. It’s all around us.

MAN: What’s the fifth planet from the sun?

HARI SREENIVASAN: Helping us search for information.

WOMAN: Jupiter is the fifth planet orbiting the sun.

HARI SREENIVASAN: Our smartphones use AI to navigate us, choosing the least congested traffic routes. Even the US Postal Service uses it to sort mail. And on Wall Street, autonomous machines help make major financial decisions.

RAY KURZWEIL, inventor & futurist: At least 90 percent of the financial transactions are guided in one way or another by artificial intelligence.

HARI SREENIVASAN: Ray Kurzweil directs Google’s engineering lab, but spoke to us in his capacity as an independent inventor. He’s convinced that AI programs are already on track to solve many of the problems vexing mankind today.

RAY KURZWEIL: They’re helping us find a cure for disease, helping us diagnose disease, analyzing environmental data to help us clean up the environment. Virtually every industrial process is a combination already of human and machine intelligence.

HARI SREENIVASAN: Large tech firms are betting big on the promise of AI Last year, Google paid $400 million to acquire DeepMind, a London startup specializing in deep learning. Facebook is raising eyebrows as it continues to pluck AI talent. And IBM is investing $1 billion to grow its Watson division, based out of new headquarters in New York’s Silicon Alley.

Remember Watson, the supercomputer which beat a pair of Jeopardy! game show champions in 2011?

MAN: Watson?

COMPUTER: What is Jericho?

MAN: Correct.

HARI SREENIVASAN: Well, in the four years since, IBM has sped Watson up 24-fold. What used to be a room full of computing machines can now fit into a pizza box, all accessed from the cloud.

You could say these are the brains that power Watson, but since all the data lives on the cloud, it’s hard to visualize.

GURUDUTH BANAVAR, IBM: What you see is how Watson works.

HARI SREENIVASAN: Guru Banavar is vice president of cognitive computing at IBM.

GURUDUTH BANAVAR: Watson has come a very long way.

We have taken some of the underlying technologies that helped us win the Jeopardy! game show, and applied it in many domains that matter, like health care, education, business decision-making.

HARI SREENIVASAN: Last month, IBM Introduced Watson Health, its entry into the personalized health care space. The idea is to use Watson’s AI to make sense of vast troves of health data to deliver tailored information to physicians, insurers, researchers and hospitals.

GURUDUTH BANAVAR: The difference between any data that previously we were able to analyze and the new data that are — we have to apply artificial intelligence techniques to is that the new data is natural language. It’s just written in English. Computers have never been able to understand natural language.

Typically, these are very high end, complex information that’s published by scientific researchers, and now Watson is able to read those.

HARI SREENIVASAN: At the Memorial Sloan Kettering Cancer Center, Mark Kris, a thoracic oncologist, is leading a team that is teaching Watson how to diagnose cancer.

DR. MARK G. KRIS, Memorial Sloan Kettering Cancer Center: We needed some way to help doctors deal with the deluge of information that’s available now.

HARI SREENIVASAN: Watson is being trained to sort through reams of information about the patient, the most current medical research, and get it to the doctor to help make a decision, all at a pace beyond humans.

DR. MARK G. KRIS: Our kind of idea here though is that this system is going to be like what we kind of call a learned colleague.

HARI SREENIVASAN: A colleague that can assist with instant diagnoses and recommended courses of treatment. The recommendations are highly personalized based on a patient’s unique genetic makeup.

DR. MARK G. KRIS: The person I’m asking about is a 55 year old man who already has had surgery for his lung cancer. It was discovered that this cancer had spread to lymph glands that were nearby.

So, the first thing this system does is, it shows all the different treatments that are recommended. And then now I ask what kind of chemo to give, and it points to a chemo regimen, two different drugs. And if I want the more information about exactly why this decision was made, there’s a little button right next to this chemo choice that takes you to the medical literature and some key publications about this regimen, the benefits it can give, and why that choice was made.

HARI SREENIVASAN: Dr. Bob Wachter is associate chair at the University of California, San Francisco, Medical School and author of a new book, The Digital Doctor.

DR. ROBERT WACHTER, University of California, San Francisco: In some ways, ironic that computers will probably be best at low level tasks, pretty simple algorithmic stuff. I have a runny nose and a cough and a low grade fever. What should I do? And high — very high complexity stuff, like, I have an unusual form of lung cancer and I have these genetic mutations, and what should I do?

HARI SREENIVASAN: But Wachter says where computers and A.I. still struggle is in the middle.

DR. ROBERT WACHTER: A lot of medicine kind of lives in that middle ground, where it’s really messy. And someone comes in to see me and they have a set of complaints and physical exam findings all that. And it could be — if you look it up in a computer, it could be some weird — it could be the Bubonic plague, but it probably is the flu.

HARI SREENIVASAN: Wachter is also concerned about fatal implications that can result from an over-reliance on computers. In his book, he writes about a teenage patient at his own hospital who barely survived after he was given 39 times the amount of antibiotics he should have received.

DR. ROBERT WACHTER: So, in two different cases, the computers threw up alerts on the computer screen that said, this is an overdose. But the alert for a 39 fold overdose and the alert for a 1 percent overdose looked exactly the same. And the doctors clicked out of it. The pharmacists clicked out of it. Why? Because they get thousands of alerts a day, and they have learned to just pay no attention to the alerts.

Where the people are relegated to being monitors of a computer system that’s right most of the time, the problem is, periodically, the computer system will be wrong. And the question is, are the people still engaged or are they now asleep at the switch because the computers are so good?

HARI SREENIVASAN: That’s one of many ethical questions facing scientists, and society, as artificial intelligence continues its rapid advance.

For the PBS Newshour, I’m Hari Sreenivasan in New York.