Two talks and a hometown

Notes from the AI in Health Conference, Houston, TX, September 15–17, 2026.

I spent the past two days at the AI in Health Conference, listening to talks on multimodal AI, healthcare, and surgical robotics. Most of what I work on lives in slides and embeddings, so it was nice to spend two days thinking a little more broadly about where AI meets healthcare. Two sessions stayed with me, for reasons that had less to do with engineering than I expected.

The ethics talk

The first was Nancy Berlinger’s talk, from The Hastings Center: Ethics in the Social Robotics Lab: What Can Philosophers Learn from Design Engineers?

The whole time, I kept thinking about my hometown.

I’m from Harbin, a city in China’s far northeast. It used to be a railway hub and one of the country’s industrial centers. When industry declined, young people left, and the population that remained grew older. Its history has some similarities to Pittsburgh’s: a city built on heavy industry, hollowed out when that industry declined, and left with an aging population and a strong medical system. Pittsburgh has largely remade itself since. Harbin is still in the middle of that story.

Today, aging is especially visible in northeastern China. People usually talk about this as an economic or social problem. It is also a medical one: more chronic illness, more people living alone, and fewer young hands available to help.

Harbin also has a sizable concentration of oncology care and research, so the city knows illness and age well. Historically, our regional cancer hospital stood next to the crematorium, and for many patients the distance between the two was too short. Medicine has lengthened it.

The crematorium has been replaced by an inpatient building. The medical campus is surrounded by hotels, much like the Texas Medical Center here in Houston, because people come from far away expecting to go home afterward.

When people talk about AI and aging, the conversation usually goes to treatment: better diagnosis, better drugs, earlier detection. But much of what an aging population needs happens outside the hospital. One example she gave is that many older people struggle with lifting, whether that means carrying groceries, moving around the house, or getting out of bed on their own. This is where robots could help.

And that is not really medical care. It is just daily life, made harder by age. It’s important to realize that, for aging robotics, we need to shift our attention to older people’s daily lives. Once a machine enters that space, the hard questions are no longer only technical. How much should someone rely on it? What should it be allowed to do? Who decides what it is for?

As an engineer, I am trained to ask whether something works and how well. The talk reminded me that there are entire disciplines built around questions beyond utility: figuring out the real need is where it starts.

The surgery talk

The second was Gregory Hager’s talk, from Johns Hopkins: AI for Surgery: Novelty, Niche, or Necessity?

This topic is close to home for another reason. My father is a surgeon. In 2006, he learned about the da Vinci system and trained to use it. His hospital began adopting it only around 2020, fourteen years later. Luckily, he saved the training video on a hard drive instead of a CD.

So before I understood much about how well surgical robots perform, I already knew how slow adoption could be. There was a trained surgeon and a system already in clinical use elsewhere, and still more than a decade passed between the training and the first real case.

Cost is part of it. So are regulation, procurement, hospital infrastructure, and institutional politics. Patients often prefer an established treatment over an advanced technology that feels unfamiliar. A new tool has to justify itself not only against the disease, but against everything already in place.

Hager drew a distinction between robots that assist surgeons and robots that automate parts of the work. Automation, as he framed it, can already handle small tasks and relatively simple procedures. The open questions are how far we actually need it to go, and where greater autonomy would genuinely help.

My own view is that assistive tools are the more feasible path for now. With a surgeon still in control, oversight and intervention are easier to preserve, and the procedure may also be easier for patients to accept. For many problems, a tool that reliably extends what a surgeon can do may be more useful than one designed to replace the surgeon’s role.

What I took home

What I took from the two talks is that need, capability, and adoption matter equally.

The important questions are not simply whether something can be built, but whether it addresses a real need, whether it works well enough, and whether people can realistically adopt it.

We spend so much time on capability, but a technology can be impressive and still answer the wrong question.

During the talk, I had this robotics idea about a colonoscopy device that could collect tissue continuously as it moves, almost like peeling an apple, for richer pathology sampling.

I paused when I imagined being the patient.