Understanding AI: A Talk for Curious Minds

AI
Data Centers
Talk
A BILL Tea & Talk: a short history of AI, a basic overview of AI today, and the role of data centers in Pennsylvania and our communities, and what we can do about it.
Published

November 17, 2026

Astrophotograph of the Eagle Nebula: reddish clouds of gas and dust against a dense field of stars.

The title slide’s photo: the Eagle Nebula, from my backyard.

Bucknell Institute for Lifelong Learning (BILL) Tea & Talk · Bucknell University · Tuesday, November 17, 2026. The audience was BILL members, mostly retired adult learners, plus community guests. Event listing: BILL special presentations.

The talk had four parts: a short history of AI, a basic overview of AI today, the role data centers play in making AI possible and the challenges they bring to Pennsylvania and our surrounding communities, and ideas for what we can do about it.

Thank you to everyone who came out, and for the great questions.

Key takeaways

1. The idea is 80 years old. The hardware isn’t.

Timeline of AI from 1943 to 2024: first math model of a neuron, the name "artificial intelligence" (1956), the perceptron (1958), the first AI winter (1969), backpropagation (1986), two gaming GPUs win an image contest (2012), the transformer (2017), ChatGPT (2022), and the Nobel Prize in Physics for neural networks (2024).

From the first math model of a neuron (1943) to ChatGPT (2022): the idea was right decades ago, and the computers were too slow until two gaming GPUs changed the field in 2012.

Sources: Dartmouth proposal (1955) · Rumelhart, Hinton & Williams, Nature (1986) · Krizhevsky, Sutskever & Hinton (2012) · Vaswani et al. (2017) · Nobel Prize in Physics 2024

2. Fluent isn’t the same as fact.

Three real cases of AI errors with consequences: a man hospitalized after ChatGPT suggested sodium bromide in place of table salt; Deloitte partially refunding an Australian government report with invented citations; more than 2,200 court rulings involving AI-invented legal citations.

A chatbot predicts what sounds right, not what is true. Treat it like a bright, eager assistant who never says “I don’t know,” and check it hardest on health, money, and the law.

Sources: Annals of Internal Medicine: Clinical Cases (Aug. 2025) · AP, “Deloitte to partially refund Australian government for report with apparent AI-generated errors” (Oct. 7, 2025) · Charlotin AI Hallucination Cases database, via HAQQ (Oct. 2026)

3. Six questions any community can ask.

Six numbered cards with questions about a proposed data center: megawatts, cooling and water source, who pays for grid upgrades, decommissioning plan and bond, where retired hardware goes, and public reporting.

Before a data center gets built nearby: How many megawatts? What cooling, and what water source? Who pays for grid upgrades? Is there a decommissioning plan and bond? Where does retired hardware go? What gets reported publicly, and when?

Sources: Governor’s Executive Order 2026-05, summarized by Morgan Lewis (Aug. 2026) · Bucknell News (Sept. 8, 2026)

4. The technology is neutral. The deal is everything.

Slide reading "The technology is neutral. The deal is everything. Ask for megawatts. Require the design. Set terms before approval, together."

Who carries the burden of a data center depends on who has a seat at the table. Ask for the megawatts, require the design, and set the terms before approval, together.

By “the deal” I mean the terms a community sets with a data center developer before a project is approved: where it can go, how much power and water it can use, who pays for grid upgrades, what gets reported to the public, and what happens when it closes. The same technology can be a good neighbor or a heavy burden, depending on those terms.

In Pennsylvania, much of that power sits with local government. Decisions about how data centers are built, and whether they fit a community, fall to township supervisors and borough councils. Zoning is their main tool, and Penn State’s Center for Energy Law and Policy notes that municipalities without a zoning ordinance “will not have nearly as much control” over data center development. Many rural townships have no zoning at all: in Snyder County, only six of 21 municipalities have zoning, and Tioga County has no countywide zoning, so development can go anywhere in a municipality without its own ordinance. Even a township with zoning can be caught off guard. In Lycoming County, Muncy Township’s ordinance was written long before hyperscale data centers existed, and when a developer applied to build three data center buildings on 100 acres of cornfields along the West Branch, the dispute went to county court, with the developer claiming automatic approval because the township missed a 60-day deadline.

That is why township boards matter so much right now. Each board should make sure its zoning ordinances address data centers directly, along with everything that comes with them: water use and water sources, land access, oil and mineral rights, power and grid upgrades, noise, and decommissioning. Each township board is responsible for negotiating a deal that protects its residents, not the data center companies.

Three Pennsylvania examples show how much the terms matter:

Money is not the same thing as a good deal. So the talk closed with three asks: get the megawatts, require the design, and set the terms before approval, together, with residents in the room.

Back to top