What exactly is an AI-Empowered University System?

And why would you empower your university with OpenAI?

Image from OpenAI’s announcement of a partnership with the California State University system. Say what you will about OpenAI, but they put nice pictures on their website.

Well, that was quick. On Sunday, I wrote about ChatGPT Edu, OpenAI’s “affordable offering for universities to responsibly bring AI to campus,” saying that “So far, there is not much evidence this attempt to “ed their tech” is getting many takers.” On Monday California State University system issued a press release announcing they will “Become Nation’s First and Largest AI-Empowered University System.” 460,000 students and 63,000 faculty and staff are a lot of takers.

Except that no one has actually received anything yet. This was a press release announcing that free subscriptions for ChatGPT will be available. When, exactly, was not mentioned. Oh yeah, and they announced a new AI website. Sorry, it is called the AI Commons Hub, "a “dedicated platform” that offers “AI tools, training programs and certifications, and CSU-developed solutions.”

As a press release, this was a brilliant success. The practice in business and tech journalism these days is to amplify hype, not ask questions. Much of the coverage acted as though this was mission accomplished, as if CSU campuses are now cybernetic meadows where teachers and students are frolicking with ChatGPT on their way to their amazing AI apprenticeships.

The reality is that it is easier to write a press release and spin up a website than it is to take steps toward responsibly adopting AI tools on a college campus. Raising a big flag that says we’re AI-empowered invites questions from faculty and students about what is happening and when. It is not clear when or if there will be answers.

Maybe CSU has all this figured out. Maybe there are teams on each campus who have already solved the problem of provisioning access for all 500,000+ accounts. Maybe there really are apprenticeships this summer for students who have been told their AI skills will make or break their chances on the job market. Maybe being AI-empowered means something more than a ChatGPT subscription and an hour-long training video about AI literacy.

But all that happened on Monday is that CSU announced its intention to hitch its 23 railroad cars to OpenAI’s steam engine. This feels sort of like the LA Unified School District’s announcement last March about its exciting partnership with AllHere to create a personalized chatbot named Ed.

This was Ed, the personalized educational chatbot. He is standing next to Alberto M. Carvalho, the Los Angeles Unified School District superintendent. Photo credit: LA Times

Maybe that’s unfair. OpenAI is a multi-billion dollar company led by a well-respected technology executive known for his honesty, not an AI startup led by an executive making promises she couldn’t deliver on who turned out to be defrauding investors. Still, the dynamic that had CSU Chancellor Mildred García making this announcement is the same one that led Alberto M. Carvalho to appear at a news conference to stand not-at-all awkwardly next to someone wearing an Ed costume.

Here is what I wrote then:

It is not hard to imagine how LAUSD got to last week’s announcement. District executives, pressured by their board or excited about ChatGPT, told their staff to get moving on this AI thing. The result is a chatbot guardrailed up to its non-existing eyebrows and trained on data already publicly accessible on the LAUSD website.

Again, the comparison may be unfair. Maybe CSU’s announcement wasn’t rushed out in a panic about the need to get moving on this AI thing. Maybe the members of CSU’s AI Workforce Acceleration Board, representing Microsoft, Nvidia, Instructure, Linkedin, and OpenAI were giving planning advice like “measure twice, cut once” and “engage your faculty and students early and often when planning major changes.” Maybe, but let’s ask who benefits and who doesn’t.

This announcement benefits OpenAI because it adds fuel to their coal car at the moment when DeepSeek had people asking if OpenAI’s engine was sputtering. It benefits CSU executives who can show their boards headlines and a new AI website. It benefits students who want to pay $20 a month for a ChatGPT subscription but can’t afford it.

Who does it hurt? It hurts colleges and university systems that are being careful in bringing AI to their campuses. Their boards will now point to the Forbes and Fortune articles and ask, “Why can’t you be like CSU?” It hurts the CSU middle managers and technologists who now have to encourage their faculty and students to use ChatGPT as a learning or research tool and figure out what the bosses meant when they said “dedicated AI platform.” You mean a website? No! I mean a platform!

The competitive pressures among the companies building generative AI products and services have created a fake-it-till-you-make-it approach that will likely end in tears for most investors. I won’t feel bad for them when the price for what OpenAI is selling turns out to be close to zero. That’s capital markets for you. Colleges and universities are in a different business, one that should reward deliberation, consultation, and transparency.

Transparency is the most interesting aspect of this announcement, something I suspect OpenAI has not fully considered. The big AI-empowered university system flag flying over CSU’s Chancellor’s office is now a target for those with questions about the value and feasibility of what OpenAI is selling. California sunshine laws will enable journalists and faculty members to answer questions like How much money is CSU paying OpenAI? How many faculty and students logged in to use ChatGPT this year? What is the actual content of the “training programs and certifications” that were promised? These are questions journalists who cover higher education and faculty senate committees will be asking.

If CSU has actual planning and effort behind its announcement, it will benefit those of us working out the slow, difficult process of implementing new technology on college campuses. We will want to know more! If this is another Ed the chatbot situation, then CSU will be a reminder that edtech startups promising an AI revolution are not good partners. There are better options than giant tech companies.

My advice to universities and colleges remains the same. Build small teams of technologists who understand LLMs and have them experiment alongside teachers and students to learn what value, if any, there is in generative AI. Try out different AI models, not just giant foundation models being created in secret. Smaller, open-source LLMs are quite powerful, cheaper, and potentially useful. If you want to sign a contract with a vendor, go with one that offers a pricing plan based on actual use, not potential use.

Most of all, remember that Silicon Valley has a big problem. They have made a giant bet that generative AI will transform society. So far, the evidence for that bet paying off any time soon is fairly thin, and investors are getting nervous. But that’s Silicon Valley’s problem. And now, it’s CSU’s problem. Don’t make it a problem for your campus.