Health IT, Hospitals

AI is “still pretty dumb” and like a “2-year-old”

Mayo Clinic’s chief information officer describes the limitations of AI describing the nascent technology as a 2-year-old child learning to interact with the world.

toddler, danger

Last month, Mayo Clinic’s CIO gave the strongest endorsement so far of artificial intelligence technology at the annual HIMSS conference in Orlando, Florida.

Cris Ross along with Tuffia Haddad, a breast cancer oncologist, at Rochester, Minnesota institution, portrayed the tangible benefits of using artificial intelligence, specifically IBM Watson Health’s AI engine.

But make no mistake.

Ross wasn’t donning rose-tinted glasses as he reviewed this emerging technology that’s set to transform myriad industries, including healthcare.

“Artificial intelligence is still pretty dumb,” Ross declared before adding, “And I don’t mean that in a really derogatory way.”

What Ross meant is the current limits of AI.

He described IBM Watson Health “as some of the best computer science on the planet” but noted that AI is heavily dependent on mammoth amounts of data. Here’s how Ross captures the limitations of AI, adding that his view of the technology may result in “fist fights”: (slightly edited)

The best artificial intelligence today is still driven entirely by so-called semantic models, which is understanding language and the relatioship of words to each other and how they build up. So the only way that these things can work is by giving them mountains of data to plow through to try and get to statistically meaningful connections, which then can be leveraged to gain some other understanding.

So, this is like a 2-year-old child just learning to speak and to walk and how they interact with the world. When I put my hand on the stove, that’s not a good outcome. It’s not something immediately clear to a 2-year-old child.

What all this AI is lacking is an ontological model where you can describe a structure abstractly. Watson had no idea what a patient was, what a hospital is, what a doctor is, what a drug is, what the effect is on a patient, what’s the relationship between a doctor, drug, a patient and an outcome.

No clue, because with these technologies you can’t describe an abstract concept and have that abtract concept be applied….

But as long as we are still based on raw horse power semantic engine technology, it means that the only place where this technology is applicable is where there is sufficiently deep and rich data sets with enough narrow variations ….

Those narrow variations allow the technology to look for some correlations and then arrive at some knowledge, Ross explained.

James Golden, senior managing director, PricewatrehouseCoopers’ analytics group, who was moderating the AI panel at HIMSS, chimed in that as AI is developed and perfected over time, interested stakeholders should keep an eye out for “deep learning.” Deep learning is a subset of machine learning where algorithms try to make sense or model abstract/thought through data. These algorithms are aimed to function as neural networks in the way a human brain does.

The IBM representative on the panel discussion at HIMSS did not resort to “fist fights” as Mayo’s Ross essentially described the best computer science on the planet as a thoughtless toddler.

“The technology isn’t the goal. The goal is the outcome, the health that we are all trying to move towards.” said Sean Hogan, VP, IBM Healthcare. “So, if Watson is still a toddler, a young infant even, glad that we’re choosing good parents or smart parents like Mayo and MD Anderson and some of the top institutions around the world and we are actively trying to learn from that experience.”

Photo: HKPNC, Getty Images

CORRECTION: An earlier version of the story misspelled James Golden’s name.