Lessons Learned About AI From A Healthcare Technology Startup Leader
Eric is President of Suki and seasoned technology executive with expertise co-founding and scaling companies including Hotwire and Expedia.
When I started my entrepreneurial journey back in the late ’90s, I was drawn to large industries ripe for disruption due to the emergence of the internet. The entrepreneurs I looked up to and admired at the time—Jeff Bezos, Pierre Omidyar and Elon Musk—brought an outsider’s contrarian mentality to the industries they reshaped. Inspired by these leaders, I spent a dozen years starting and scaling travel companies before shifting my focus to home furnishings about 10 years ago.
In the back of my mind, I always was fascinated by healthcare due to its sheer size and broad societal impact. After all, no other industry affects all of us as deeply and directly as healthcare, which is a big reason why healthcare comprises almost 20% of the U.S. GDP and more than $4 trillion in annual global spend.
Healthcare is also a deeply personal industry in spite of its size—full of one-on-one interactions between clinicians, patients and a seemingly endless variety of administrators. The personal nature of healthcare brings additional issues like data privacy, patient rights and legal liability to an already complex industry. If entrepreneurship is largely built on learning an industry in order to transform it, the initial learning curve for healthcare is steep and daunting.
A few months ago, I found a talented team of entrepreneurs crazy enough to tackle some important challenges in healthcare, and I’ve started the process of getting up to speed on a new industry, just like I did in travel and home goods. So far, my personal crash course has made healthcare’s massive size and complexity even more apparent, and it occurred to me that others can benefit if I write about my learnings and observations.
Let’s Start With Artificial Intelligence
There are so many thought-provoking elements of healthcare to write about, but I’d like to start with artificial intelligence (AI), which stands squarely at the intersection of technology and healthcare and represents a significant opportunity to both increase the quality of patient care as well as reduce costs. AI isn’t just a buzzword but rather a broad swath of computing technologies designed to replicate and speed up human tasks. Machine learning (ML) is one particularly noteworthy AI technology, which can be described as the use of complex statistical models to analyze massive amounts of structured and unstructured data to quickly produce insights that would take humans a very long time to digest and deliver on our own.
As consumers, we’ve seen how AI and ML enable Big Tech to leverage data to anticipate what information, products or services we might need. Era-defining businesses have been built on the back of AI and ML, from Google Ads to Amazon’s Alexa. AI needs data like humans need oxygen to grow and reach their full potential, and although the vast amount of personal information now collected by Big Tech may be unnerving, there’s no denying that the resulting services have become an indispensable part of our digital online life.
Current Concerns And Uses Of AI And ML In Healthcare
AI and ML applications to improve healthcare hold much promise, but results to date have been somewhat mixed, as some existing algorithms have been found to deliver inconsistent or even flawed results. In other cases, what’s presented as AI is, in fact, something more pedestrian, for which humans behind the scenes are providing the mental muscle (often in offshore locations).
That said, improvements in database management, ML and digital imaging present compelling reasons for optimism. Although we’re still in the early stages, there have been successful applications of AI in specialties like radiology, for which researchers have developed ML algorithms that leverage the superior visual analysis capabilities of advanced computers to identify potential cases of breast cancer that human doctors might miss.
Steady increases in computing power now also allow companies like Google to make diagnoses using techniques impossible for human doctors to duplicate. For example, Google now uses advanced ML to comb through massive datasets of patient data, including vital signs, smoking history, aging patterns and even retinal scans to deliver highly accurate risk estimates of cardiovascular disease that physicians can’t match.
In other situations, an ML-based assessment based on clinical data can inform a doctor’s decision-making. For example, advanced ML can now handle many of the calculations used in cardiac MRI environments and other specialties so radiologists don’t have to perform the computations manually, which results in accelerated clinical diagnoses with higher levels of accuracy.
Now Come The Questions
These breakthroughs, although exciting, raise even more questions in an already complex industry. Is it possible for AI and ML to enable a higher quality of medical care in underserved areas with fewer doctors? How can AI and ML be overseen and regulated if humans can’t re-create the approaches used? Who’s to blame if an algorithm is wrong and leads to tragic consequences? Longer-term, are AI and ML a supplement to human doctors or a replacement?
These questions are already starting to be partially addressed and answered, as the FDA has recently approved some algorithms for clinical use without physician oversight but, in these cases, the sponsoring company assumes legal liability for any errors. Imagine a future courtroom battle where the defendant isn’t a doctor but an algorithm. Not exactly high drama but, who knows, perhaps an enterprising screenwriter will create a new antihero by merging Hal from 2001: A Space Odyssey with Jack Nicholson’s Colonel Jessup in A Few Good Men.
Stay tuned as I pass along more observations and learnings as I further educate myself on all things healthcare. Buckle your seat belt, as we have a long journey in front of us!