In the last month, the public discussion around Artificial Intelligence has escalated dramatically, with researchers, technology leaders and lawmakers openly questioning if increasingly powerful AI systems could pose a serious risk to all of us — and what government should do about it.
It is true that without proper management these incredibly powerful machines may have the potential to cause harm. And governmental oversight to mitigate risks is essential.
But it is also true AI can be an extraordinary scientific tool — and that we have a responsibility to make sure those tools are available to researchers working in the public interest.
Part of the problem is the label itself. “AI” has become so broad that we are collapsing fundamentally different technologies and concerns into one debate. The recent Navier-Stokes breakthrough is a good example: AI may have accelerated an important scientific discovery, while also raising serious questions about how quickly frontier systems are advancing and who controls that progress. Both are “AI” stories, but they are not the same story.
Realizing the potential of advanced AI for the public good requires access to these systems, which is typically only available at a relatively small number of private companies.
But researchers in New York are now using the most powerful academic research computer in the country to better understand disease, develop new medicines, model climate risks and solve scientific problems. In other words, some of AI’s most consequential uses are happening far from the consumer marketplace.
I’m talking about Empire AI. It was created in 2024 under the leadership of Governor Kathy Hochul, who brought together some of New York’s leading research institutions to put powerful computing back into the hands of academics. Today, we are a nonprofit consortium providing scientists with computing resources most individual institutions simply could not build on their own.
We’re not building consumer products or chasing revenue. We are providing researchers with an extraordinarily powerful scientific instrument and asking what they can discover with it.
As just one example, at Weill Cornell Medicine, researchers are using Empire AI to improve prostate cancer diagnosis and treatment. The work requires analyzing complex biological and clinical data that no human could process on their own, and would be difficult to manage even with traditional university computing resources. The goal is to help doctors identify cancer more accurately, tailor treatments more effectively and, ultimately, improve outcomes for patients. And in all likelihood, those patients will never know their improved care was “powered by AI.”
And that’s just one project. Since its launch nearly two years ago, Empire AI has assisted more than 1500 researchers across 471 projects. Already, the machine has worked for about 768,000 GPU hours. And every hour on Empire AI would have taken these same researchers sometimes days or even weeks on computer systems that are typically available to academic researchers.
This is the promise of Empire AI: not to put “AI” into more of everyday life, but to give scientists a more powerful tool for answering questions that matter to everyday life.
And this distinction matters when we talk about risk. Using advanced computing to analyze medical data, model proteins or simulate extreme weather is fundamentally different from developing increasingly powerful general-purpose AI systems. These are different technologies being used for different purposes, and they do not present the same set of concerns.
But taking AI seriously also means taking its risks seriously. That’s why New York is not only investing in a powerful computing tool, but also independent academic research into the technology itself.
At Binghamton University, the new Center for AI Responsibility and Research will focus on foundational research into the safety, security and transparency of artificial intelligence. The center was established with support from New York State and a major philanthropic effort led by Bloomberg co-founder and Empire AI board chair Tom Secunda. Its mission is to help researchers understand how these systems behave, where they fail and how they can become more secure, transparent and worthy of public trust.
Establishing that trust is an important part of the model.
Right now, much of the conversation around our work is understandably about the infrastructure — the computers, the facility and the extraordinary scale of computing power we are making available to researchers – as well as the possible risks. My hope is that several years from now, that won’t be what people are talking about. I want our success to be measured by what researchers discover, and whether those discoveries make people’s lives better.
Kiran Keshav is director of Empire AI, New York’s nonprofit consortium providing advanced computing resources to academic researchers across the state.