Skip to main content

Mike McSherry, CEO and Co-Founder

Is Bing really making a comeback against Google? How did that happen?

Practically overnight ChatGPT and Microsoft’s investment in them changed the market. AI is powering Bing, Microsoft Office and even Microsoft’s Nuance division is partnering with Epic for voice scribing. 

In response, Amazon rushed a move with Hugging Face that resulted in a dip in stock price. Feeling the pressure, Google made a similar move with comparable results. In the space of a month, Bing gained legitimacy. 

As a digital healthcare innovator, why do I care? Because the pervasive cloud wars have morphed into AI wars, and they’re coming for healthcare. New fortunes will be gained and lost in terms of market share, positioning, and value.

These are exciting times. AI is the new black, and the buzz around it will benefit developers and eventually our healthcare system clients. However, we must be careful not to put the bot before the horse.

The company OpenAI is open sourcing its AI (ChatGPT), which means enterprise customers can create their own AI initiatives on their large language models (LLMs), and AI comes up with the most interpretive data sets. LLM combined with AI produces the largest, most groundbreaking insights. Now that the AI is more readily available, who wins or loses the race to AI dominance will come down to who has the largest and most unique data sets for AI to interpret.

Who has the largest data sets in the world? Healthcare.

Who stands to benefit the most? Healthcare.

What started as a down-the-road obsession among wonks like my colleagues and me is now a full-fledged thing. Kudos to OpenAI and its ChatGPT for popularizing AI tools among the masses. When new technology achieves pop-culture relevance, we gain from the momentum in every industry, including healthcare.

That said, we can’t help but notice the hype coming from some healthcare marketers – and there is a lot of bluster.

Many healthcare claims about AI simply aren’t true – and much of it isn’t ready for prime time. As much as I want the claims of AI and Machine Learning to come to fruition, this is healthcare, not consumer tech. Mistakes don’t lead to inconvenient or confusing consumer experiences, mistakes in healthcare can be disastrous.

Having said that, where are we going?

Cloud wars set the stage

It all started here. Initially, cloud’s most important use was for storage. When storage prices dropped, GPU processing (compute) of that data became the focus. This is why GPUs are important. In addition to large data sets, AI requires fast, powerful processing to compute and run insights.

Computing strength matters because those with it can quickly compute AI interpretations.

Looking at this, Oracle’s ambitions in healthcare make sense, as they are arguably in a strong position to win. The company was the world’s biggest database company – all on premises. Then, cloud became standard and on-prem, along with its limitations, fell out of favor. Azure, AWS, and Google each have larger cloud market shares than Oracle. 

To improve Oracle Cloud Infrastructure and grow its market share in healthcare, Oracle acquired Cerner. Today’s war leverages cloud market share to power AI-driven insights derived from cloud data models – and every single company has its own proprietary LLM.

The power and possibility of AI grow as the data increases, and the technology is best leveraged with vast data sets to run machine learning models. These are objective facts. In healthcare, we see successful and robust AI implementations in radiology and oncology because of their ever-growing data sets.

However, we need more robust deployment and wide-scale adoption across the digital health ecosystem to harness the full scope of AI’s capabilities. Not enough vendors can aggregate sufficient data.


Current uses and opportunities

The brilliant researchers at MIT and Massachusetts General Cancer Center recently unveiled a new oncology AI oracle.

Known as Sybil, it calculates the probability that an individual will develop lung cancer within the next six years (check out the study). The researchers applied the tool to patients with no smoking history to catch cases that had previously slipped through prescreening cracks (FYI, nonsmoker diagnoses currently are double recent decades’ totals). 

AI also takes a patient-convenience approach for current and former smoker diagnoses:

The guidelines recommend that those over 50 undergo a yearly low-dose computed tomography (LDCT) chest scan. Still, fewer than 10% of that group do so, according to the researchers — so the ability to predict up to six-year cancer risk from a single scan could help improve current diagnosis rates.

Now we’re talking about saving more lives, improving efficiency, and maximizing ROI. Great stuff.

There are several practical benefits to be pumped about on the radiology side. The medical imaging industry anticipates the widespread deployment of AI to help augment techs and radiologists. The expected benefits are expedited exam throughput, retake reduction, image quality improvement, dose reduction, speedier organization and data pulls, autocomplete of structured reports, and automated measurements.

A bright future

Expect the biggest AI implementation successes in life sciences and propensity modeling.

Even as venture capital in healthcare languishes, we see bright spots specific to the growing intersection of biotech and computation. One trio recently managed to raise $350 million – an oversubscription of $50 million – for their new venture capital company, Dimension.

Regarding propensity modeling, the possibilities continue to be heady to us. Look for social determinant and lifestyle choices to be incorporated into these models. According to recent estimates, an individual’s health outcomes are influenced 40-50% by behavior, 20% by physical and social environments, and 30% by genetics. That means we’ve barely scratched the surface of many data sources.

The healthcare industry’s primary challenge regarding successful AI implementation is its unique fragmentation and variability. Larry Ellison, Oracle co-founder and CEO, went on record during an acquisition presentation last summer about healthcare’s data shortcomings and the need to address them.

He promised a nationwide architecture to solve longtime interoperability challenges and enable real-time data access (h/t Healthcare IT News).

“Your data is scattered among a dozen or two dozen separate databases, one for every provider you’ve ever visited … We’re going to solve this problem by putting a unified national health record database on top of all these thousands of separate hospital databases.” 

Health systems should adopt the “buyer beware” mentality when it comes to proposed AI solutions for the time being. We just aren’t there yet, and once we are, decision-makers should still question the data models and the use of AI toward effectual output and ROI measurement.

Lay your AI foundation now

Is your health system collecting all the data it will need when AI ramps up in earnest? Assume your competitors are.

  • Don’t let overpromising third-party vendors be the ones in charge of your patient’s health data
  • Break down data silos
  • Create a governance structure for managing current and future data
  • When AI’s power and necessity become more evident, make sure your health system is ready and experiment now, especially with back office automation and manual processes. Clinical recommendations should, naturally, be treated with more caution.

AI will play a significant role in healthcare—eventually. We at Xealth are well-versed in its possibilities and are building our technology toward a future that embraces it. 

Still, we promise never to lay claim to leveraging AI until we have enough data to run the statistical, machine learning, and natural language processing (NLP) models reflective of actual AI-driven solutions.

Close Menu