Authored by Alex Kotlar, PhD, Founder and CEO, Bystro AI
Gary Vaynerchuk said something recently that I haven’t been able to shake.
His basic argument was that 10 years from now, regurgitating information won’t be worth much. AI will be able to do that faster, cheaper and probably better than we can. What becomes incredibly valuable instead is emotional intelligence.
Gary has been early, and right, about enough shifts in the digital world that the idea is worth taking seriously. From brands on social, to influencers, to digital collectibles. But it also gets really interesting when you apply it to medicine.
We’ve spent the last few years debating whether AI will replace doctors. I think that’s probably the wrong debate. AI is going to replace a lot of what doctors currently spend their time doing. That’s different.
Think about the average visit to a primary care physician. You explain what’s wrong. They ask questions. They look at your history. Maybe they order bloodwork. They take all of that information, combine it with what they learned in medical school and what they’ve seen over years of practice, and try to figure out what is happening.
There is obviously enormous skill involved. But there’s also an enormous information-processing component.
And AI is getting very good at that part.
I see this firsthand because we’ve spent the last several years building Bystro AI around one of the most complicated information problems in medicine: genomics.
A single person’s genome contains millions of genetic variants. Researchers trying to understand those variants have to navigate enormous datasets, scientific papers, population databases and highly specialised analytical tools. Until relatively recently, asking a seemingly simple question about DNA could require a geneticist, a bioinformatician, specialised software and a fair amount of time.
That’s a hell of a barrier to entry. Bystro (www.bystro.io) is trying to knock it down.
We’re building technology that lets researchers interact with genomic data using plain English. Instead of knowing exactly which database to query, which software package to use or how to write the code themselves, a researcher can ask a question and have the system help interrogate the underlying genomic information.
That’s not about making scientists less important. It’s about making their expertise go further.
And once you spend enough time working on that problem, you start looking at the broader AI and medicine debate a little differently.
Physician and author Robert Wachter has suggested autonomous AI could become health care’s “economy class”: cheaper and more accessible than seeing an AI-assisted physician, but ultimately a lesser product.
A recent JAMA Viewpoint pushed back on that assumption and entertained the possibility that autonomous AI could eventually outperform physicians, even AI-assisted physicians, on some cognitive medical tasks.
Maybe it will. Maybe it won’t. I’m more interested in the person who doesn’t have either option today.
What about the patient who lives three hours from the specialist they need? The person who can’t get an appointment for six months? The community hospital that doesn’t have a geneticist on staff? The researcher sitting on valuable genomic data without an army of bioinformaticians available to analyze it?
Giving those people access to expertise isn’t “economy class.” It’s giving them something they didn’t have before.
That’s democratisation, and I think we’re badly underestimating how important that could be.
But here’s where Vaynerchuk’s point makes this even more interesting.
If AI makes medical knowledge dramatically more accessible, the doctor’s job doesn’t disappear. The value of the doctor shifts.
Your AI may eventually know every medication you’ve taken, every lab result you’ve ever had, your family history, your genome and every relevant medical paper published through yesterday afternoon.
That’s all well and good, but it still has to deal with you.
As humans, we don’t always do what the data says we should do. We get scared. We procrastinate. We misunderstand risk. We don’t take the pills. We don’t change our diets. We hear “95 percent survival rate” and somehow only hear the other five percent.
Good physicians understand that.
In some ways, general practitioners already spend an enormous amount of time managing the psychology of medicine. They reassure. They persuade. They explain the same thing three different ways. They figure out when a patient needs data and when that patient simply needs someone they trust to tell them, “This is what I think you should do.”
That’s emotional intelligence. And it may become significantly more valuable as the informational advantage physicians have historically held becomes less scarce.
We’ve already seen versions of this transition in other industries. The internet didn’t eliminate financial advisors because stock prices suddenly became available to everyone. Expedia didn’t eliminate travel advisors because people could book their own flights. Google didn’t eliminate teachers because students could look things up themselves.
The jobs changed because access to information changed. Medicine needs to change too.
At Bystro, we’re betting that highly specialised scientific knowledge shouldn’t remain trapped behind technical barriers simply because that’s how the system has always worked. Genomics is our starting point because the gap between what the science can tell us and how difficult it is to actually access those answers remains enormous.
Large language models give us an opportunity to shrink that gap.
There are plenty of reasons to be cautious. Medical AI needs to be validated. Answers need to be grounded in real data. Hallucinations aren’t cute when somebody’s health is involved. Privacy and security aren’t optional. And there are decisions where a qualified human absolutely needs to remain in control.
Those are engineering, scientific, regulatory and ethical problems we have to solve. They aren’t reasons to preserve artificial scarcity.
The future I find interesting isn’t one where AI replaces every doctor. It’s one where a researcher without a bioinformatics department can interrogate genomic data. Where a community physician can access expertise that once lived almost exclusively inside major academic medical centers. Where patients can understand more about their own biology. And where doctors spend less time functioning as extraordinarily expensive information retrieval systems and more time doing something machines may never be particularly good at: understanding the human sitting across from them.
Maybe Gary is right. Ten years from now, knowing the answer may be the cheap part.
Knowing what to do with a human being once you have it could be the most valuable part of medicine.
