Part 1: Why We Need Data Infrastructure…Just Not Everywhere
Building Smarter, Not Harder...
Photo taken from my old porch in Thomas at sunset on July 28, 2023
Intro to this series can be found here.
Notes from a Mountain Motel
A cotton candy sundown sky dance has started as I find the only late night internet around at The Billy Motel in Davis, WV. No complaints here, as a late May chill has hit the mountain and a futuristic MCM fireplace warms me while I work. From my seat, I can look up from my computer screen to enjoy the view of the woods leading to the Blackwater Falls state park.
When I first moved here, the massive Blackwater Falls was my distant echoing late night, front porch stargazing soundtrack. Although five miles away, on very quiet and very cold nights, it would drift down the slope into town. Unfortunately, those same stars, echo and sense of peace are threatened to be erased from accessibility.
We Need Digital Infrastructure
But...Before I get too deep in the woods of this series, it is important for me to acknowledge on of my biases and why I have it. To preface that I am a firm believer in the need to develop digital infrastructure…within reason. It’s so central to my life that even talking about data infrastructure, digital divides and ethical AI or AI safety sparks a deeply felt fire of passion in my core.
I am a certified nerd.
I aim for techno-realism.
Despite my zeal for the building blocks of our inevitable future, I believe it’s vital we do it better than the lack of democratic, environmental and human rights values that you’ll read about in this series. It’s also why I want to tell my view of this story as someone who sees multiple perspectives on the issue of digital development.
The Work That Informs My View
For 6 years, I have been researching tech ethics, its social and environmental impacts, as well as how we (humans) communicate about technological development and the stories we build to help us adapt to changes we didn’t ask for. I have an MBA and MPA in Sustainable Solutions with a focus on tech policy and deradicalization. I’ve read lots of books, listened to hours upon hours and hours and hours of tech, AI, and global policy podcasts, lectures, and conversations, including those on the original iteration of Clubhouse.
***Side Note: RIP OG Clubhouse. Iykyk. While most of the rooms I took part in were focused on global politics, cybersecurity, emerging tech, (and I was an active member in a fake cult called Prelon) there were rooms for everyone on this app. If you want to know more, here are two articles about the good things that took place during the brief popularity of the app. Like this one here about the use of Clubhouse to promote free speech in the Middle East, and here about how it became a support platform for educators during the pandemic.***
I’ve also been fortunate to have shared podcast space with two incredible co-hosts and some inspiring guests for a season of TechDirty2Me. Season 2 will be released by the end of the year.
My current research fellowship is with the Digital Impact Alliance (DIAL). DIAL is focused on digital public infrastructure (DPI). ***None of the opinions I express in this series are theirs and insert here all other legal qualifications that free them from any liability or unintended implication.***
DPI technical definition:
DPI refers to the foundation of digital systems, (i.e. identity verification, data exchange frameworks, and payment platforms) that enable secure, interoperable, and inclusive delivery of services. DPI makes life easier for everyone by developing frameworks that reduce friction in everyday interactions, especially when it comes to those under the care of governing bodies.
One example:
If you’re in a state, like *cough* Ohio *cough*, dealing with the DMV means multiple visits to multiple departments across multiple days. It’s annoying to those who have the privilege of time, but even worse, it’s an unsustainable burden to those with accessibility restrictions in time, finances and physical resources.
All of that stress could be eliminated through a secure platform, accessible by phone, with identity verification operated through a public version of something like Plaid, from the comfort of your couch. Or while you’re eating lunch at work.
A system where you could transfer your ID info across a digital border from one state’s DMV to another, flawlessly. Reducing the chances of a DMV employee scoffing, “how can I know this is you?” at an out-of-state license with your picture on it, while looking right at you, as a valid form of identification even though that other state obviously already vouched for your identity.
Also, a system that makes accessing DMV public assistance like having trafficking violations waived if you’re in financial duress easy. As it should be.
Apply this ease of access to all government social services. Developing those pathways is just one part of DPI. And if you’re someone who cares about people having accessible care from government services, then you should care about DPI.
Thing is, all the data that’s collected in the process of offering better digital social services has to be stored somewhere. Which means, not all data centers hold evil.
Data Representation Matters
It used to be that we had boxes and boxes and files and books and cabinets stashed in spooky storage centers. The likes of which most of us have only seen in crime dramas. Those storage spaces, that have held everything from unsolved crimes to historical census data to social service records, have been known to catch on fire and flood year after year.
Some of that data is held on site, meaning in computers and small servers within the institutions gathering them. Some is sent to larger data centers. Either way, more agencies are making digital copies of physical files as a means of backup, and all that data gets stored somewhere. Businesses do the same thing. The larger the business or the higher up in government you get, and the larger the storage needs are.
Nuance is Important:
Now, I know the data center in Tucker County is very likely to be used to fuel AI. The argument I'm making here is that not all data centers need to be avoided indefinitely. What they’re used for really matters. Same as how they’re managed.
Regulating AI is one way to potentially slow down an overhyped data center boom in America. AI regulation is a worthwhile pursuit. However, keep in mind that stopping the whole AI train would take regulations on a scale that are unlikely to be adopted within the next few years. Which means data centers built to support AI will still be developed to some extent.
***Side Note: This is one area where the palpable tension between AI Safety and AI Ethics means they really need to just kiss already and make a baby. The friction between those two crowds is too much to get into here. This essay by Professor Olle Häggström goes into more detail if you’re interested.***
Do we need all these data centers to develop a technology we don’t fully understand the implications of? The answer to that question depends on where you fall on the techno-utopian-solutionism to techno-dystopian-catastrophism spectrum.
Despite these times being referred to as a data center gold rush by some news sources, the industry might actually be seeing early signs of a bust, partly due to supply chain challenges, among other issues, like the need to consume massive amounts of available natural resources.
Global Disparities in Data Infrastructure:
Again, I’m not anti-data center in all contexts.
I’m pro human-centered, sustainable development.
Currently, the United States has over 5000 data centers. By far, the most out of any country in the world.
South America has around 400 data centers. Africa has less than 300.
THESE ARE CONTINENTS!!!! The two continents combined have less than a 5th of the data centers of the United States.
Most of my current research revolves around global south digital development, specifically how governments can learn from the mistakes and best practices of others to better manage their data, specifically dark data (unused/unanalyzed data), and data storage in a pro-social and efficient manner. It has been an eye-opening and humbling experience to become more and more aware of rest of the world outside of my American mind.
As I’ve experienced even in well-meaning progressive circles: most Americans, myself included, have not truly explored the world and therefore do not understand how our activism here can often result in perpetuating colonialism when we apply narrow, albeit well-intended beliefs, to other countries without their input.
Even the “woke” are mostly still asleep when it comes to understanding global realities because the lack of lived exposure leads people to assume problems are the same everywhere.
The Missing Voices in AI development:
We need data centers…
…in underrepresented areas.
As it stands, data collection is tragically lacking in the global south which means all of the emerging AI systems are being developed with the worldviews of the global north who have more digital access. What assumptions can we then make from knowing most AI is trained on data from countries with certain backgrounds and worldviews? Seems like a reasonable question to ask, IMO.
…We don’t need AI colonialism.
AI is only as good as the data it’s given.
When it’s missing huge swaths of the global population, that means its views and understandings are still very, very limited.
Digital access still isn’t even a guarantee in the global north! This is largely a class issue, where those without means of access are left behind in digital storytelling that’s shaping the next iteration of data-powered services like AI.
Data centers are like the pages of a book the story is being written on. And yes, they currently amplify environmental impacts beyond that of the historical book industry, but technological development isn’t going to stop which means we need to come up with better solutions than saying EVERYTHING needs to stop.
Because this train isn’t stopping.
So let’s ask, who is on this train and where is it going?
One thing is for sure, we need the stories of those that have not yet been collected to be part of the pages of history. Without their voices, our picture is incomplete. We’re missing other perspectives that could truly help us see the world through a healthier lens.
Great…now what?
Data centers aren’t ideal. Obviously.
They consume a disturbing amount of energy and water. Disturbing in part because how we supply them means the resources used are ruined, stained, and made toxic by their use. The byproducts are an infected version of the original.
So what do we do?
We adapt.
We look at the problems that have come up and we say, “How will we make this better?” Because if the train isn’t stopping, and we’re wasting time wishing it would stop, it’s up to us to build the tracks further down the line to steer it in a direction we want it to go.
Also, I’ll address some practical “Possible Solutions” in a few posts from now.
First, I need to get back to the story of Tucker County and what we can learn here about what NOT to do, before we can look at how to do it better.
Here’s one suggestion to start:
Some areas should be protected. Some should always remain wild.