End of Summer Plus Why AI Discussions are Complicated AF
Hello, hello! Last week, I spent a couple of days in the Windy City to celebrate the end of summer and had a great time. Visited Molly’s Cupcakes—one of the bakeries featured on Cupcake Wars which... If you've ever watched Cupcake Wars... triggers a need to either bake or consume bakery or both. I love being able to visit actual eateries and locations featured on reality TV shows. Hoping one day to drop by The Repair Shop, too. Wound up visiting the Shedd Aquarium; it was a perfect experience, despite all the remodeling. Had the opportunity to pet a sturgeon and had a fun tail-less stingray encounter.
Of course, no end-of-summer celebration is complete without imbibing the Herald of Autumn’s preferred potion—a Pumpkin Spice latte. Yes, I had one. No, I have no regrets. Tastes like sweater weather to me.
When I got home, dived into Genshin Impact and explored Natlan. The music is so well done—just listen to a video of Natlan’s theme song performed in Swahili.
Now that I'm settled, I’m all set to continue moving forward on several fronts. I’m happy to share more journalling prompts next week, too. This week, I owed you a random topic. So, on that note... I’d like to present my Grand Unf*ckening of the Discourse around AI.
Why AI Discussions are Complicated AF
What exactly is AI, anyway? According to New Scientist, “Artificial intelligence or AI simply means software used by computers to mimic aspects of human intelligence.”
Let me repeat that very important word: mimic. Think about this with respect to writing or art. AI doesn’t teach the end user how to write or how to draw, because it doesn’t “know” how to write or draw. It’s simply programmed to search its datasets to mimic content human beings have already created.
The reason why this conversation is complicated is twofold: AI does not have a standardized definition and its applications are not singular. Calling everything “high tech,” “powered by AI,” “smart,” etc. is frustrating, because there’s a lot of room to confuse consumers for different purposes—not all of which are nefarious. Often, when you read a definition, there’s a slant that recontextualizes AI. Analytics website Tableau, for example, states that: “Artificial intelligence is a specific branch of computer science concerned with replicating the thought process and decision-making ability of humans through computer algorithms. There are many different branches of AI that can create and do different things. Some types complete simple tasks, while others are much more complex. Some AI programs adjust their own algorithms, and some specialized algorithms are so advanced they can beat human experts in their given fields.”
Mimicking aspects of human intelligence is, semantically, different from stating that AI replicates the thought process of human beings. New Scientist’s definition infers AI cannot replace human beings while Tableau’s does. And that’s just two definitions—there are far, far more—before we begin considering the applications of AI that includes, but isn’t limited to, algorithms, predictive text, content generators, data analysis, chatbots, etc.
Without a standard definition, it can be hard to confer guidelines about its usage and efficacy. There are questions I do think we should be asking. Questions like: “Where has the data been sourced from to train AI?” “Do those datasets require constant upkeep?” “How does AI affect creators?” “Are there hidden costs of AI that we’re not yet grasping?”
The answer to the first question is that the data must be collected from a “source” in order for it to function. As labelyourdata.com points out, you can draw from Open Source, or copyright-free materials, the “Internet”, and Artificial Data Generation—which is also copyright-free. The data on the Internet, however, is not necessarily copyright-free nor is it free to use however you wish. Go back and read those Terms of Service you scrolled past. Or, open up most creative works—they’re not copyright-free, either, but they’ve been treated as such by for-profit companies who need the data in order for their software to function. Just because content is freely available to be consumed, doesn’t make it “free.”
What’s more, that functionality is limited. A machine’s deep learning has diminishing returns. Unfortunately, in order for AI to continue functioning, continuous data collection, typically the kind of data generated by human beings, is required because that “learning” is pattern analysis. University of Illinois-Chicago states that: “Artificial Intelligence (AI) enables machines to learn from experience, adapt to new inputs, and execute tasks resembling human capabilities. By leveraging AI technologies, computers can undergo training to perform particular tasks through the analysis of extensive data sets and the identification of patterns within the data.”
At present, every creator whose works appear in digital form runs the risk of their human creations being used to train software that mimics their finished piece—regardless of copyright and permission.
The cost to creators should be obvious, but just in case the illicit data collection isn’t enough to convince you? Across multiple creative industries, creators have been replaced by AI because the software is believed to do as good of a job writing, illustrating, etc. as a human being. Except, that’s not entirely true. What is happening is that writers are being fired, then one or two writers are rehired to edit the works produced by AI—and don’t get any credit for doing so. This actively hurts a writer’s ability to build and maintain a porfolio, because they cannot point to their byline. When the AI takes the credit and is attributed as the author, this devalues the craft of writing and skews the idea that AI is far, far more effective than it actually is. For authors, in particular, there’s also been a dramatic uptick in AI-generated books appearing in online stores—which floods the eBook market and erodes the reader’s trust in buying from authors they either haven’t already read or haven’t heard of. As such, this makes it much, much more difficult to discover new readers to sell books.
Though I do believe AI has its uses in broad-scale data analysis and similar calculation-based tasks, the mimicking aspects of the software poses a significant risk to people who create. In fact, any job that doesn’t require human interaction is at risk of being replaced by AI provided companies can not only justify the cost—but also figure out how to continually update and improve their datasets. The predictions are dire.
For myself, I’m most interested in two aspects of this discussion: the impact on creators, which I’ve addressed somewhat, and the environmental concerns. AI is technology, and that technology often requires server centers that consume substantive amounts of power to stay running and water to keep them cool. Water, however, is not an unlimited, renewable resource, especially given that water used to cool server farms condenses and evaporates. This usage might be mitigated for one server farm, but what most people are missing is the scale of the issue. Forbes stated that: “Already AI’s projected water usage could hit 6.6 billion m³ by 2027, signaling a need to tackle its water footprint.”
6.6 billion cubic meters. Hoo! Big number, right? Well, the problem with evaporating water isn’t exclusive to AI, because as temperatures rise around the world, water heats up. The combination of a number of factors means that there’ll be far, far less water in the near future. (I’m also considering other vectors. At present, seventy percent of the world’s freshwater resources are used by the agriculture industry. When the population rises, that specific figure also increases, because crops require water to grow.) I know 6.6 billion cubic meters is hard to wrap your head around, so think of this: The CIA World Fact Book has an overview of water usage by country. Pop over to that link. Notice anything? The projected water usage to cool server farms in the United States eclipses the total water usage in several countries. Take, also, into account the fact these server farms either are (or have already been) constructed in municipal areas around the U.S.. The more industral water usage increases, the greater the chance for evaporation, the less water people in and around bigger cities will have access to.
Besides the impact on water, power grids, etc. there’s also another aspect I haven’t seen discussed regularly with respect to the data centers: the demographic of the people affected. There have been articles touting how AI is changing rural america, despite the fact that not every company discloses where their data centers are being built and maintained. Business Insider, The Observer, and other outlets openly discuss the locations when larger companies disclose the affected areas. “Who” would be affected by a water shortage?
That answer, dear Reader, is going to have to wait for another time and a lot more data collection and analysis of my own. I do, however, believe that we haven’t fully grasped the ethical and environmental considerations, how AI in general will affect the nature of work, and what it actually costs us when compared to its positive, data-crunching attributes. For creators, however, I do think the answers are far more clear.
Next week, I’ll be announcing my schedule at Game Hole Con and September’s journalling prompts. Until then? Keep on, keepin’ on.