Based on the limited IQ of humans, can artificial intelligence (AI) be trained to be smarter than people?
Is human emotion more important than IQ? Is it possible for humans to train artificial intelligence (AI) to be smarter than people? Our expert panel weighs in:
A Healthcare Analytics News™ Peer Exchange®
David E. Albert, M.D.: John, one of the issues is, today, when we’re talking about the great strides in AI, we’re talking about the great strides in multilevel, deep neural networks. That’s the technology that has made Google Photos able to say, “find my wife on the beach,” and boom picks up the pictures. Or translate my Alexa, you know: “Alexa, what time is the football game this afternoon?” That is a pattern-recognition technology.
Yes, a lot of artificial intelligence is: I remember you, you’re John Nosta; I remember you, you’re Kevin Campbell: that’s pattern recognition. But that’s not everything involved in intelligence — we’re going to see whether or not it can become creative. I’ve heard about art that’s been created by deep neural networks. I’ve heard about music that’s been created. I’ve heard about writing. And we have companies [as] sophisticated as Google who say they’re AI first, or Microsoft AI first, and they’re going to push the boundaries. I’m waiting to see if they can do more than just win at “Jeopardy” or at ”Go.” And we’re going to find out. I don’t know.
>>Watch the last episode: PX: Is AI a Physician's Tool or Future Physician?
Geeta Nayyar, M.D., MBA: To add on to that, right, not everything is a pattern or an algorithm. I’m a rheumatologist. There’s not enough data on certain rheumatic diseases.
There are certain populations, we don’t have enough South Asian women with lupus to even study, what is the best treatment? So, it’s not that simple. We have to remember that these are tools. These are absolutely tools, just like we have always seen in medicine, whether it’s robotic surgery, whether it’s going from open-heart surgery to less invasive interventional radiology tactics and procedures. The evolution of medicine is this story. Whether today it’s AI and tomorrow it’s something else, I think it’s phenomenal, the opportunity is phenomenal. But it’s not the silver bullet that everyone seems to think it is. It’s not the doctor that’s the tool. And the same for the patient. The patient’s not the tool either.
Kevin R. Campbell, M.D.: Well the patient’s the center of the universe for what we do.
Geeta Nayyar, M.D., MBA: The patient is the center and the key. And oftentimes you have a patient that tells you, “I just don’t feel well, something doesn’t feel right.” And you as a doctor start doing your investigation, especially in rheumatology.
Kevin R. Campbell, M.D.: It’s also a balance, and this is very trite, of IQ versus EQ [emotional quotient]. AI has a high IQ and zero EQ. I think that doctors are being built now to have higher EQs, acceptable IQs, but [EQs] may be a more desirable position.
David E. Albert, M.D.: That’s certainly a noted venture investor, Vinod Khosla’s, view, who famously said, “80% of doctors will be replaced by AI.” But what he did say is that we’re training, but maybe we should be, instead, training and cultivating more empathy into our doctors. At the same time, we should be training doctors who are also quantitative.
>>Watch the fourth episode: Machine Learning: The Future of Medicine?
John Nosta, B.A.: Or maybe we change the game and have IQ/EQ and TQ — maybe there should be a technology quotient. Why do we live in the bounds of humanity? I find it very, very sad that we hold humanity as the standard upon which every innovation is based. Why can’t we be smarter, better than humans? Why not?
Kevin R. Campbell, M.D.: Because we are human beings. And as human beings, we are emotional creatures, and that’s part of what makes life worth living: hurting, crying, laughing, enjoying company, going through a health journey with a patient—that’s what makes medicine worthwhile for me.
John Nosta, B.A.: But our sensory perspective is meager, Kevin. You know our ability to hear an opera is limited; a dog hears an opera better than we do; we process it differently. Our hearing, our eyes, our smell are meager compared to the broad electromagnetic spectrum that we bathe in every day. I’m just saying that there’s a broader reality. And somehow, just saying, “Well it’s as good as a human,” it’s just not; it’s not the end game I want to play to.
David E. Albert, M.D.: You’re right, but remember what I brought up about big data, well-labeled data — so who labels that data? Today, it’s experts labeling the data. So, do they make mistakes? I can tell you: I just had three board-certified cardiac electrophysiologists review 2,000 EKGs [electrocardiograms]. Two of them are young; young meaning less than 40. And they disagreed on 350 of them. And then I had a guy my age, who’s very well published, adjudicate. So, you have three experts; they’re all well trained, academic and incredibly well respected, and they disagreed 15 to 20 percent of the time. And so, you know I worry that no matter how good our data [are], it’s still going to have errors in it, and that means it will train.
>>Watch the third episode: The Promises and Pitfalls of Big Data in Healthcare
John Nosta, B.A.: You know, yeah, you’re right, you’re spot on, but does that —
David E. Albert, M.D.: But you can do pictures and say there’s a cat in this picture, or there’s not a cat in this picture. And that data [have] shown to be very effective at training deep neural networks. When you have rheumatology — my wife’s a rheumatologist and understands your problems — that’s a very complex notion. You have all these autoantibodies and you have all these symptoms.
Kevin R. Campbell, M.D.: I can’t even keep up with them all.
David E. Albert, M.D.: It’s unbelievable. At the same time, you have a patient coming in saying, “I hurt every day. I can’t do this can you help me?” And so, I think, I believe that AI is going to be a critical tool. But I also think, you know, you’re right, maybe it’ll be smarter than us, but we’re the ones training it. So, I have a little bit of a problem of thinking, if an expert is training it, maybe it’ll be smarter than most of us.
Get the best executive insights directly to your inbox. Sign up for our daily newsletter.
FDA: Software Defect Prompts Synaptive Medical to Recall 3D Imaging Surgical Tool
AI Struggles Across Health Systems, Needs Wide Range Testing​​​​​​​
How AI Is Shaking Up Healthcare, Beyond Diagnostics