Amid Data Center Chaos, What Can Regular People Do?

I messed around with a communication ledger for data center projects over the weekend and hit quite a few pitfalls. Recently, I saw a news item on IT Home where US Treasury Secretary Besant criticized AI companies for not explaining data center matters clearly to communities. Trump warned that opponents would eventually fall behind and become poor. My first reaction wasn't to pick a side, but to ask: can we build a small tool to see if these projects explained themselves clearly?

First, let's define two terms. A data center is a facility housing servers and cooling equipment. Community communication refers to informing residents about impacts on electricity, noise, water, and land rent before construction begins. Besant said the industry "hasn't explained its business at all."

Beginners should look at this first. We aren't building a system; we're going from 0 to 1 creating a spreadsheet, collecting a few projects, and separating resident objections, company responses, and final statuses.

Day 1: Just collect. Open a browser and search "data center opposition [State Name]" or "data center community opposition." Find news, open Excel or Google Sheets (online spreadsheets), click File > New, and write column names in the first row: Project Name, Location, Source, Resident Objections, Company Response, Current Status. Don't categorize columns yet; just copy the original text. Expect one row per project.

I recorded a pitfall. On Day 1, I wrote "involving approximately $150 billion in investment" from the news into a single project's notes. By Day 3, I realized this was the total investment involved in multiple shelved projects, not the amount for one specific project. Later, I added a "Data Scope" column to record only the original text's range.

Day 3: Start letting text generation models help organize. Text generation models are programs that read a passage and write summaries, like ChatGPT. Open the webpage, copy the original news text, and input: "Based only on this text, extract resident objections, company responses, and project status. Leave blank if not written." Then ask it to output three columns. The result should be short sentences, row by row. If it starts hallucinating, change the prompt to "No speculation, no words not present in the original text."

Here's a pitfall. Initially, I asked the model to summarize the entire news article, and it mixed corporate responses with community complaints. Later, I split it into two steps: first extract objections, then extract company responses. Use --- to separate them in between.

After a week, the ledger becomes useful. You need at least five projects from three different sources. Then add a "Communication Failure Signal" column. Examples: Only project introduction, no impact on electricity bills; Only corporate website, no community meeting records; Residents repeatedly ask about noise, response only mentions technical upgrades; Hearings scheduled late at night or during weekday daytime. The problem might not be that residents don't understand AI, but that companies aren't speaking human language.

Don't rush to connect APIs. If you want automation, register for model services and get a string of keys. This key acts like a password for the program, allowing the spreadsheet to call the model. Feed news summaries to the model and ask it to output JSON. JSON is a table machines can read. Beginners are advised to stick to manual spreadsheets first.

In my tests, the easiest method is a fixed template:

Please output JSON with fields: project, location, opposition, company_response, status, source. Write null for fields with no information.

Copy the result and import it using spreadsheet tools. The first import often fails due to Chinese commas, quotes, or line breaks. Change to English commas, remove extra blank lines, and try again.

The last step is writing a judgment for yourself. Not "Should data centers be built?" but "Did this project answer the cost concerns most important to the community?" Besant said the AI industry explains things terribly; I agree halfway. The industry needs to lay out numbers for electricity, water resources, land use, and job opportunities. What ordinary people can do is turn a pile of news into a comparable table.

After learning this, try exporting the ledger into a simple bar chart to see if objections concentrate on electricity prices, noise, and land. It's okay if the chart is ugly, as long as you can see which type of response is the scarcest.