I am working with an intern this summer on a project evaluating how universities are using artificial intelligence. We are each researching schools close to us. She has Loyola Marymount University and the wider California system, and I am working through the New York schools, currently comparing NYU and Columbia. We have both done our initial research, and we are spending the second half of the month thinking through what we found and working it into an essay. To get to that essay I need to practice, so I am writing short posts like this one to work through the ideas one at a time until they hold together.
The first glaring difference I found between NYU and Columbia was not a policy or a syllabus, it was a posture. Columbia is investing in building their own models, standing up open-source LLM chat interfaces, and developing their own tooling around AI. They are also diversifying their use of commercial services, leaning on Google, ChatGPT, and Claude rather than picking one. NYU, from what I can tell, has gone all in on Google. One campus is building internal capacity and keeping several vendors in play. The other has bought a stack and handed the keys over. Both of those are defensible on a spreadsheet. Only one of them leaves the institution with something of its own at the end.
That difference is the thing I keep circling back to, and it is why I picked this topic for the summer in the first place. It is not really a university question, it is an enterprise question wearing a university uniform. If you are not investing in internal capacity when it comes to applying artificial intelligence, you are outsourcing your core competency, and you are giving away most of the value your organization produces in the process. You do not learn what your own data is worth. You do not develop the people who could have told you. You do not build the muscle that lets you switch vendors, negotiate, or say no. You end up renting your own thinking back from someone else, and the price of that rental is set by them, not you.
In the case of a university, the stakes are different in kind, because whatever posture the institution takes, it is also teaching that posture. A school that builds its own tooling and keeps multiple providers in play is showing students that these systems are things you can open up, evaluate, and choose between. A school that pipes everyone into a single vendor’s suite is showing students that AI is weather, something that arrives from a company and that you adapt to. Instead of empowering students to make their way into the world with these tools, you are demonstrating how to be managed by them, and how to hand your agency and your knowledge over to the handful of corporations behind AI today. Students absorb the shape of the thing they are handed far more than the policy language wrapped around it.
There are plenty of other ways NYU and Columbia are using AI that I want to work through, positive and negative, and I will take them one at a time on the way to something more thoughtful. My intern Emily has been finding a version of the same split out west, particularly with UC Irvine investing in their own technology rather than renting all of it. I am staying focused on the east coast, but I am learning a lot from her research, and I am genuinely looking forward to her opinions on what matters here, because she is closer to the receiving end of these decisions than I am. There is a lot we can learn from students, and a lot students can learn from those of us working in business sectors already being reshaped by this. This exchange of ideas across generations and coasts has been the best part of my summer, and it keeps pulling me outside the AI box I work inside every day.
