Johns Hopkins already leads the medical world. On AI, it won’t be easy.

The surgical assistant slid a long slender tube up the patient’s nose, a camera and light helping guide the removal of a tumor from the tightest of spaces behind the eyes.

“Move left, please,” the surgeon told his assistant holding the device known as an endoscope. “Move right.”

This was only a test; the patient wasn’t real.

Neither was the assistant.

It was a humanoid robot, named Sprite by students, that represents one of Johns Hopkins University’s top priorities for the next decade: artificial intelligence.

Hopkins, already a global leader in medical research, is now recruiting researchers and students to its AI tech hub.

By harnessing the power of AI, Hopkins hopes to develop new drugs, diagnostic tools and administrative services. The goal is to improve the health care system and patient care, and overcome the flaws and unease that come with relying on algorithms.

Sprite’s AI-powered brain has been trained over many months by Johns Hopkins University researchers to move with subtlety and stamina unmatched by humans.

Eventually the robot could assist in hours-long procedures without a peep, relieving staff and expanding patient access to care in places without enough surgeons, said Mathias Unberath, the Hopkins computer science professor leading the project.

”A robot won’t complain,” he said.

AI is a big swing, with big potential gains for the university and state eager for spin-off companies and high-tech jobs. Hopkins officials estimate an $800 million impact by 2029.

But AI’s rapid development is already raising significant concerns.

Some cities have begun pushing back against the power-hungry data centers needed to fuel AI, citing financial and environmental impacts. That includes Hopkins’ one in East Baltimore designed to support the university’s controversial AI institute.

There are broader fears about how AI could jeopardize patient privacy. And some studies show AI’s output is often just wrong. Observers and experts warn that overall, society’s guardrails lag the pace of the groundbreaking technology’s rapid deployment.

“It’s going to be an amazing period of innovation and change, the likes of which most people have never seen,” said Dr. Derek Angus, a professor at the University of Pittsburgh School of Medicine who led a major 2024 report on AI in medicine.

“Hopkins isn’t alone,” he said. “It’s really exciting — and terrifying.”

The Johns Hopkins Data Science and AI Institute was launched in 2023 to bring an array of projects under one roof. It’s likely among the nation’s largest, expected to add 110 faculty jobs over five years.

More than 1,000 students and graduates will also train inside in that time, becoming the next generation of AI pioneers.

Hopkins and other U.S. universities still have ground to make up. By one measure, U.S. News and World Report, for example, ranks Chinese universities as having the best academic research performance, which could mean that any technological and economic spinoffs from those institutions would remain in that country. Hopkins trails at No. 63.

Competing will take a huge pot of money. Officials declined to provide an overall budget for the institute or identify all of its specific funding sources.

But some spending has been revealed. The $196 million East Baltimore data center includes $9 million in state funding. On its main campus in Homeland, Hopkins will also spend up to $10 million on street upgrades around the new institute’s two new academic buildings expected to be completed in 2029.

Much of the research will still rely on the federal government. But that has increasingly become a less safe bet. Last year, under the Trump administration, federal grants for new and ongoing research at Hopkins fell by 43% compared to the year before — a loss of more than $500 million.

To fill the gaps, financial analysts say foundation and corporate grants, venture capital and partnerships with big tech companies are playing a larger role. For example, Nvidia, the American tech company now worth trillions of dollars, has helped Hopkins develop Sprite’s capabilities.

Mark Dredze, a computer science professor and director of the Hopkins AI Institute, believes that despite funding challenges, universities will continue to play an essential role as incubators and training grounds, and even policymakers.

“Guardrails are critical,” he said. “The first thing we need to think about is who is looking into those guardrails. Universities have a critical role.”

Corporations are usually proficient at building products and making money, but according to Dredze, some may be less adept at assessing risks, prioritizing social needs and preserving the public‘s voice in the process.

In 2017, Unberath moved from his native Germany with his wife and two young children to the United States. And in that move, he saw a chance to join a community committed to innovation and work on broad public acceptance of technological advancement.

He’s already making progress on a bevy of new AI applications.

Unberath helped develop an AI-powered smartphone app to help consumers diagnose the common bacterial infection strep throat at home. Another invention in development: a tool to assess people most likely to develop glaucoma, a leading cause of blindness.

He helped develop an early warning system to detect the potentially deadly hospital infection sepsis that was just approved by the U.S. Food and Drug Administration. It’s one of more than 1,400 AI-powered tools approved so far by the FDA, the majority in radiology.

To keep the human perspective in mind, Unberath regularly conducts research to see how people interacted with AI tools and how the technology could gain acceptance.

Sometimes, it could come down to fashion.

For Sprite, he’s examining what the robot might wear in the ER waiting room to engender trust among patients and their loved ones: scrubs, a lab coat, nothing?

He also probes the algorithms powering AI innovations for bias and other flaws. He found one in a model that was only detecting skin cancer if there was a ruler nearby. That’s because the model had been trained on images that included a ruler.

Another AI project at Hopkins is a novel effort to explore how computer programs might help screen children in the emergency room who are at risk of gun violence and then intervene as needed.

Today’s health care system relies on doctors and others who are spread thin, which to Unberath raises the question, “Can we delegate work to an algorithm?”

The challenge, he said, is whether they can rely on the automatic result. “We have to know if it’s bullshit.”

A landmark summit of researchers, big tech companies and others on AI in 2024 hosted by the Journal of American Medical Association found little consistency in assessing if AI tools were working as intended.

The summit‘s participants called for more standards to help ensure enough testing before the tools developed in places like Hopkins were used in the marketplace, as well as continued monitoring once they become widely available to consumers.

Angus, the lead author of the summit’s report, said in an interview that building this consensus would take time and likely new legislation. States such as Maryland have begun passing piecemeal laws, but federal lawmakers have yet to pass a uniform measure.

“This revolution is unprecedented; we’ve never had something so large and widespread and fast as AI is going to be.” Angus said. “We need to ensure there are incentives for responsible development — and not just fast development.”

Dr. Masaru Ishii is a head and neck surgeon who is working with Unberath’s team to train Sprite the robot, an off-the-shelf model made in China. He said proving its abilities could take years. And the effort may partially eclipsed by similar research in other competing countries.

Still, he said, the effort was critical because of how often surgeries today rely on human-held endoscopes.

Ishii sees a future where specialized “kits” from a library are plugged into AI-enabled robots, giving them new skills immediately. He likened it to a scene in the movie “The Matrix,” where a human instantly learns to fly a helicopter.

“As robots get better,” he said, “the surgeries a surgeon can do become more diverse.”

Unberath’s team, with Ishii help, will continue training Sprite, using videos, virtual reality headsets and actual practice surgeries on dummies and cadavers. Real surgery could be a couple of years off.

But Ishii added “rarely do we see something where we get so universally excited.”