Bad news for IBM's super computer "Watson" -- it's not helping cure cancer. In fact,it might be making some cancer patients worse.
Watson was a bid deal in 2011, when it was on the TV quiz show Jeopardy, and beat the greatest Jeopardy champion of all time, Ken Jennings. Clearly, if a computer can beat someone at trivia, it can cure cancer. So that's what IBM announced Watson would be doing.
I have to say, I was pretty excited about it at the time. I wrote about Watson in 2012, and again in 2013.
The plan, as I said in 2013, was to give Watson 1.5 million
patient records and outcomes; 600,000 pieces of medical evidence, and 2
million pages of text from medical journals and clinical trial records. Watson could then use all of that data and apply what it learned to new patients, and recommend treatment.
It didn't work out as they had hoped. According to The Wall Street Journal, there are no published research reports that show Watson has improved patient outcomes. In some cases, the computer didn't do anything that doctors couldn't do. In other cases, Watson recommended the wrong treatment.
On the plus side, some say that consulting with Watson is helping doctors stay up-to-date.
But a computer is only as good as the information that it is given. that's part of the problem, which is kind of ironic -- the humans that feed Watson the data about cancer aren't fast enough to keep up with advances in cancer.
(That's kind of cool -- we're learning more about cancer than the computer can keep up with.)
I wish I could say that I predicted that things would go this way, but I didn't. I was excited about it all, six years ago.
I'm still excited about it. There's probably a place for Artificial Intelligence in cancer care. But probably as part of a team that includes humans.
If nothing else, humans are probably better suited to pick up the kinds of subtle cues that point to problems with a patient's mental, emotional, or spiritual health. There are some AI programs that can help people with mental health issues. But ideally, an oncologist would get to know a patient well enough to know when there are issues.
(And I did say that in 2012 and 2013, because I'm awesome.)
Of course, just like a super computer is only as good as the information it is given, and could only pick up problems it is told to look for, a human doctor can only solve the problems he or she knows to look for. A doctor who doesn't bother getting to know a patient isn't going to pick up on those cues. And doctor who don't recognize the importance of a patient's emotional and physical health won't even ask about it -- no matter how super he or she is.
I say this as I start looking for my 5th oncologist in 5 years.
I don't want a super computer. Just someone who will listen.
Subscribe to:
Post Comments (Atom)
2 comments:
Thank you for the update, I was really excited at the news of Watson helping with cancer at first as well.
And what about Joe Bidn’s “Moonshot for Cancer” or whatever it was called?
I haven’t heard word one about that in recent news.
As always, thank you for your blog.
Donna
Donna, the Moonshot initiative is holding a national summit on September 21. I think we'll hear some updates then. As far as I know, the project was funded for 7 years, and they've been quietly chugging along.
Post a Comment