You say 'silo' as if it were a bad thing...

Information storage in the AI era

Everyone seems to want to break down silos. They are monsters to be destroyed. In consultant-speak, silos stand for isolation, inefficiency, and a myopic refusal to collaborate. The “silo mentality” is a dysfunctional reluctance to share information. In universities, silos are the cause of duplicated efforts, internal competition, and the failure to innovate to serve the customer-student.

Everything about this view is wrong. In the AI era, the responsibility of researchers and specialists to preserve domain-specific knowledge with integrity, to keep going deep while others go broad, will be more important than ever. The universities that protect disciplinary silos are the ones who will thrive.

In fact, the metaphor makes my case. A real, physical silo is a marvel of functional design. Silos store and protect grain. Silos preserve quality and integrity. Without silos, grain rots. A silo is a sanctuary. Its purpose is preservation, not isolation. Here’s a picture:

Brief history of the silo (before getting to AI)

Much of human history is the story of catastrophic agricultural losses. The invention of the silo in the nineteenth century reduced grain losses from 50% to just 2%. Silos transformed farming from a seasonal survival struggle into a year-round productive enterprise.

There had been attempts for millennia to store grain (think Joseph and his storehouses in Egypt,¹ most likely underground). In Europe and America, traditional storage was underground. Beginning in the decades after the Napoleonic Wars, European agricultural scientists began experimenting with new models.

Perhaps not coincidentally, the silo in the U.S. was born in the 1860s and 1870s, the same decades that saw the establishment of the nation’s great new research universities. Fred L. Hatch, a 22-year old University of Illinois graduate, built the first model in fall 1873 at his father’s dairy farm in Spring Grove, Illinois. He had studied under Professor Willard Flagg Bliss, who had translated European scientific papers on silage. Hatch’s silo measured 10×16 feet with an 8-foot pit extending 16 feet above ground, lined with rocks and mortar, like a basement wall.

Neighbors watched as the Hatch’s dairy cows munched through the winter on the preserved corn silage. They produced more milk. Silos! Soon everyone was building them.

In 1891, a Cornell-educated professor of agricultural physics at University of Wisconsin-Madison, Franklin Hiram King, developed the King Silo, and previewed it at Wisconsin's Agricultural Experiment Station. This is the cylindrical design that became the template for modern tower silos.

How do they work? The fundamental engineering principle is the creation of anaerobic conditions to slow grain respiration and inhibit harmful microorganisms. That is, silos prevent oxygen seeping into the grain with airtight sealing systems. Glass-lined steel silos use breather bags that function like lungs, expanding and contracting with temperature changes while preventing air infiltration. Carbon dioxide injection and nitrogen-modified atmosphere packaging create environments hostile to aerobic spoilage organisms.

Before silos, life was measured in losses. After silos, farmers stored grain during harvest-time price lows to sell during peak periods. Silos enabled year-round milk production. Silos enabled strategic grain reserves for communities, creating buffers against seasonal shortages. Rural economies stabilized as community grain elevators enabled everyone to work together.

The agricultural silo is in fact one of history’s most transformative innovations, solving the storage challenges that had given farmers a headache for millennia. Silos are one of the fastest agricultural technology adoptions in history. From Hatch's single silo in 1873, there were 50,000 by 1895, and over 500,000 silos dotting the American landscape by the early 1900s. Today they are long, rather than tall, using the same technology but in plastic bags.

Back to the metaphor

The function of the academic discipline is not unlike an agricultural silo: it is a structure of preservation and quality control. A discipline preserves the accumulated knowledge, methods, and standards of a particular field of inquiry. It provides the deep, specialized training and methodological rigor necessary to produce expert knowledge. You cannot have a breakthrough in quantum mechanics without first mastering physics. You cannot develop a new gene-editing technique without a deep grounding in biology. Not everyone is qualified to tinker with the genetic code. Not everyone is qualified to perform surgery or design nuclear reactors.

Silos are good. Leaving grain out to rot is bad. When you consider the agricultural analogy seriously, silos are great. Without the disciplinary structures that ensure quality control and systematic development, intellectual progress remains precarious.

The criticism is that you can’t share and collaborate easily with silos. This isn’t true. Disciplinary experts in silos choose all the time to collaborate with those in other silos. Ask them. The Manhattan Project brought people trained in silos together. Then most went back to their silos, with joint appointments in new units. No silos were destroyed.²

The attack on “silos” usually comes from people outside of a silo, generalists who don’t have deep disciplinary knowledge or focused training. These people don’t want their ideas validated by a community of experts. They find expertise to be inconvenient. The image of the silo as narrow, contained, a kind of ivory tower, seems to support the claim that those in them are narrow, out of touch, or secluded. Someone who has never experienced a famine might easily dismiss grain storage as unnecessarily protective.

Workforce Demand vs. The Ivory Silo

When legislators want a curriculum that meets “workforce demands,” what I hear is politicians calling for grain storage techniques based on market surveys. It turns out the workforce alignment debate is old, and involves silos.

Alan Marcus’s “The Ivory Silo: Farmer-Agricultural College Tensions in 1870s-1880s” (1986) is about this very battle 150 years ago, which pitted farmers against agricultural colleges for failing to serve their practical needs: too much irrelevant coursework by faculty disconnected from “real” farming challenges. Faculty, including those in the new emerging discipline (silo) of agricultural science, wanted to transform farming from what they viewed as backward, tradition-bound practices into a modern, scientifically based profession. This group believed that only through rigorous scientific training could agriculture advance and compete with other industries.

The agricultural scientists were split – some wanted a purely science curriculum, others wanted French and German (to read papers on the latest advances in Europe) as well as Greek and Latin, because everyone should be well educated. The 1890 Morrill Education Act (known mostly for requiring states to either open their public university’s doors to everyone, regardless of race, or establish HBCUs as Land-grant institutions) gave both the farmers and agricultural scientists a win. The Act established federal oversight (so farmers had a say) while legitimizing scientific approaches to agricultural instruction. Even as the agricultural scientists could point to the Franklin Hiram King silo after 1891, emerging, as it did, from university science, nothing was resolved. The same battles are raging today.

The work of knowledge production is the intellectual equivalent of feeding the world. Yes there needs to be a workforce. But there also needs to be researchers working in silos, inventing silos.

Silos in the AI era

Before the AI era, the best defense of academic disciplines was that they are essential to protecting knowledge from political rot and political winds that result in intellectual famine. You can make this case from the left or the right.

In the AI era, disciplinary silos are crucial. First, they are a bulwark against hallucination. The rigorous, peer-reviewed knowledge produced within an academic silo is our insurance policy in the AI era. We can double check facts. Academic silos are the best source of how exactly discoveries were made and validated. The scientific method, historical source analysis (historiography), statistical validation, and other methodologies are developed, taught, and enforced within disciplinary silos.

Second, disciplines create the depth that makes AI breadth most useful. One of the recognized benefits of LLMs is their breadth, drawing connections between disparate fields in ways that might take a human years to uncover. But breadth is shallow and useless without the ongoing digging, complex theoretical modeling, and focus of researchers in disciplinary silos.

Third, academic silos are where a great deal of new knowledge is being made. Sure there will be new knowledge created by AI systems. But academic silos currently are the best sites of debate and self-correction, where old paradigms are challenged and understanding is refined. This dynamic process is essential for ensuring that future AIs are not trained on permanently outdated or biased information.

Bottom line: if you tear down the silos, do not expect the harvest to improve. Without proper storage, abundance becomes scarcity. Societies that cannot preserve their knowledge face starvation. Hunger follows decay.

1. Genesis 41:48-49: ^“And he gathered up all the food of the seven years, which were in the land of Egypt, and laid up the food in the cities: the food of the field, which was round about every city, laid he up in the same. And Joseph gathered corn as the sand of the sea, very much, until he left numbering; for it was without number.” Here’s what archeological scholarship suggests about the design.

2. Following the Manhattan Project's conclusion in 1945, key scientists returned to their traditional academic departments while continuing to collaborate. J. Robert Oppenheimer maintained his physics foundation and became Director of Princeton's Institute for Advanced Study (1947-1966). Enrico Fermi returned to the University of Chicago Physics Department while establishing the interdisciplinary Institute for Nuclear Studies. Edward Teller maintained physics appointments at the University of Chicago (1946-1952) and UC Berkeley, while directing Lawrence Livermore Laboratory.

The institutional innovations that emerged from this experience (national laboratory partnerships, interdisciplinary institutes, and hybrid research structures) were additions to existing academic frameworks, not replacements. No traditional academic departments were dissolved or merged. Instead, parallel structures were created to enable collaboration while maintaining the specialized training, peer review systems, and methodological rigor of disciplinary departments. The Manhattan Project’s success depended on disciplinary training.