How might the representations and discourses of the “AI Forward” university be placed within broader conversations and positions on the near future (4 years) of Frontier AI. This is a first effort at mapping out these positions, with all the various caveats. I’d appreciate your thoughts.

What is the current situation?

Here is one useful visualization that informs you if AI is profitable yet. (As of writing on Monday July 13th, it was $1.5T spent and $769B in revenue. And in the time it took me to transcribe that data, another $8M was spent. So, no: AI is not profitable yet. And while the topic of an AI bubble is broadly discussed, the market remains bullish, as they say. Firms such as Gartner and Goldman Sachs project $8T+ investment over the next 4-5 years.

Needless to say, it is a staggering investment. It is not hyperbole to describe it as a civilizational infrastructure project. You might ask an AI what else one might do with $8T. A 25-year global war on terror is one possibility or thereabouts ($5.8T on military appropriations and $2.2T on veteran obligations according to Brown’s Cost of War Research Series). The sheer scale is audacious.

Market Messages

While the overall market message remains positive, there are some dissenting voices inside the firms as well as far more critical journalistic and academic voices. The less bullish to slightly bearish internal positions are cautious about over investment. While, in their view, AI is inevitable, that doesn’t mean that people and companies trying to build it won’t fail or that investors won’t make poor decisions. These positions all also see high “option value” for the data centers that are never used by AI if the bubble pops. That is, they believe these data centers will still have other value to other people that will help to mitigate the losses.

The capital expenditures the comprise much of the bubble aren’t only data centers. There are new power plants, electrical grid upgrades, and fiber and network installations. Those could be mostly good, though they might be in the wrong places now. And people still do not want them built nearby. Nevertheless, there are other parts that could have more option value than computers and chips specially designed for inference.

Meanwhile the investigative/journalist media space gives us more critical voices. For example, Ed Zitron represents one of the most skeptical voices on the financial aspects of AI. He is skeptical of option value in Frontier AI’s capital expenditures to say nothing of the value proposition of LLMs themselves. In short, he is warning of more dire consequences. He argues here that we should not bailout AI companies when (not if) they fail.

Other interested corporate/national parties

Undoubtedly firms that have invested in AI have something at stake, but there are other interested parties:

• Consulting firms such as McKinsey and Gartner that advise corporations on AI strategies.
• AI chip designers and makers such as Nvidia.
• Neocloud corporations that sell compute
• Hyperscalers (Google, Meta, Microsoft, etc.) who provide cloud infrastructure.
• Frontier AI labs that buy from both neoclouds and hyperscalers.
• US and other governments who view AI as a security concern, weapon, and organizational tool.

While not necessarily asserting maximalist claims of AGI super intelligence, they are validating a civilization-scale infrastructure project.

University Map

In the consideration of university positions on AI, it is worth noting that 10 years ago, there would have generally been two groups of faculty researching AI. On the hand there was computer science, engineering, linguistics, and cognitive science building AI systems. On the other hand there was STS, media studies, philosophy and law engaging in critical analysis of AI.

Critical AI/STS Scholars: I would place myself among this group (in case the rest of the blog didn’t make that clear). There is no central position here except for a shared skepticism for the claims made by Frontier AI. That said, if this is unfamiliar, you might start with Critical AI. For a more public intellectual version of these arguments, Cory Doctorow is quite visible.

Educational and Organizational Transformers: These would be faculty with a generally positive view of AI. They believe the problems are fixable and that AI will transform both learning and our institutions in positive ways. My experience with faculty who hold these views is that they primary come from business, educational technology, or learning sciences. These are disciplines are among those who have increasingly adopted AI as a tool as the technology diffuses into workplaces and cultures. Ethan Mollick is a salient example. His bullishness recognizes the usefulness of AI as a productivity tool. Mollick’s co-intelligence suggests working together rather than job replacement or Doctorow’s reverse centaurs.

Academic Competitive Imperative: In addition to universities sensing an institution-wide imperative to respond to AI, access to computing infrastructure has become integral to remaining competitive in AI research and in a growing range of computationally intensive disciplines. Faculty who develop AI systems remain concentrated primarily in computer science, computer and electrical engineering, statistics, applied mathematics, robotics, and computational linguistics. These faculty are often positioned—and position themselves—as the foundation for institutional AI literacy initiatives.

Academic Maximalists: less a disciplinary position than one held by individual faculty. Interestingly, unlike industry maximalists, many academic maximalists, often in or close to computer science, see the incredible power of AI, but they have concerns. Geoffrey Hinton and Nick Bostrom are examples.

And the “AI Forward” University?

AI Forward universities, including my own, represent a mix of the transformative and imperative views within the academy. They also call on the positive possibilities of the maximalists to shape narrative but rest on pragmatism for addressing problems. However, they follow along with the overall AI cultural and market bullishness. I would argue that much of the academic transformer’s enthusiasm is a reflection of the AI cheerleading in workplaces encouraging workers to use it. Indeed some of the units on my campus (not Arts and Sciences to my knowledge) have designated cheerleaders. I think most faculty have encountered encouragement to use AI on the job. Similarly the competitive imperative faculty come from the same disciplines as AI researchers working building these models. It is not surprising they share perspectives on the potential of AI.

These are not criticisms of those positions. Instead, they are observations that these positions are not independent of those held in the corporate world by professionals with similar academic profiles. I would suggest they are reinforcing. As such, I think it is fair to say that the excitement that drives market investment in frontier AI drives university investment too… literally in dollars from the state, donors, research grants, and corporate partnerships. This excitement and investment also drives the shape of university curriculum, research, and academic engagement with artificial intelligence.

In short, the AI Forward university is financially and structurally committed to a bullish investment in frontier AI. Like the rest of us, they are at risk of losing investment portfolio value if the bubble pops. However, they have also made longer investments in curriculum, departments, tenure-line hires, and capital projects. And they have made reputational investments. Unlike investment firms that always warn you that your money is at risk, investments in universities should be more reliable. But they aren’t anymore.

What happens to AI Forward universities if/when the bubble pops?

Obviously it will depend on how big the bubble is. At an outer limit, my sense would the panic of 1873. It lasted 4-5 years and was called the Great Depression until 1929. As always there were a number of factors such as over investment in railroads, currency issues with silver coinage, and wars. (Fortunately we don’t have any comparisons to those factors like cryptocurrency and still more wars.)

However it doesn’t have to be that bad to bring financial and reputational damage to the AI Forward university. All that really has to happen is that decreased confidence in investment in Frontier AI destabilizes the value proposition of investing in the AI Forward university. The financial damage will be obvious in all the money spent transforming the university to be AI forward. AI-centric hires. Direct investment in AI technologies. The construction of AI departments and degrees. That could easily be tens if not hundreds of millions in wasteful, irresponsible investment. (Or at least that’s how it will be described at that point.)

That said, the reputational damage might be much worse. Inside universities, relations between and among faculty, staff, and administrators have already been strained by AI. What I have experienced is that many bridges were burned to push the AI Forward university vision. I don’t imagine the bridge burning is over either and may worsen if the bubble begins to loom. How could faculty possibly respond to the administrators that have pressured faithfulness to the university AI vision?

For students and others who rely upon the reputability of the university, these failures will only reinforce growing skepticism about academics in America. If universities can be so easily taken in, what does that say of them? For students who are in the midst or have already completed these recently assembled AI themed degrees, what has become of the value of their education? Did they realize that when they were investing in their professional future they were also gambling on the stock market?