A notebook of nutgrafs

My thoughts on AI are complicated

Existing on the internet has felt like sensory overload lately.

Scrolling on social feeds is an endless exercise of "is this AI?" and reading comment sections full of AI accusations. Everywhere I look, people are talking about AI.

Nobody asked, but I've been meaning to put my thoughts about AI in writing.

As a note: I wouldn't be sharing my views here if I was unwilling to budge on them. I expect not everyone will agree with this. I am genuinely interested in your opinions, so please send me an email or write a post in response to this if you'd like.

My worldview (AI aside)

In order to remain sane, I've established a thought framework that more or less shapes my outlook on life and what I ethically/morally/politically subscribe to.

First, the most important work happens in your backyard.

Hopelessness is the feeling I get when I see another headline about crooked politicians doing crooked things yet again. People yell about it online, but they keep getting elected.

Journalists tend to get stuck in echo chambers that make us think that everyone knows what is going on in our communities. In college, it's especially easy to fall in that trap when you are surrounded by journalism professors, news writers and fellow students. But every once in a while, we're reminded of the reality: Most people really, really do not know what's going on, whether it's the local, state, national or international scale.

Although this is rare, I think it is perfectly acceptable to pay no mind to national and international politics in favor of knowing what your city council gets up to. In my work, I see the immediate effects that local legislation has on community. Most of the way that government gets involved in our lives is at the municipal level.

Critically, this is not to say that I don't personally engage with politics beyond the local level. I'm civically engaged β€” to a somewhat excessive level, sometimes β€” and I wish everyone else was. But that isn't realistic.

From a previous post of mine I just linked above:

As I see it, my life would be pretty miserable if I let every depressing event in politics get me down, because a whole lot of it is bad.

This sounds a bit doomer, but hear me out. I believe having only grand aspirations for society leads to feeling that all efforts are futile.

Yes, world peace is a nice idea. We should aspire to inch toward it.

But we won't get there. Especially not in my lifetime, and likely not before Earth turns into a ball of flame.

I don't see a point in worrying about issues the individual cannot change when there are things we can change in our immediate communities. Personally, I feel like it keeps my head on straight to volunteer locally when I have the time. There is a demonstrable impact (in my case: helping build a community garden, packing lunches for low-income schoolkids, watching an "aha!" moment in a kid who is learning something new) and it often introduces you to people that make you feel better about where the world is headed.

I am from the American South. I do not like our politics a lot of the time, but some of the most diverse, progressive and vibrant communities call this region home.

This is not to say that I think it is wrong to try and have a larger impact. It's important to get angry about things we should be angry about, but I think the crux of my belief is that we can't let that anger control our lives.

Context about how I use AI

I bring my ideological framework to my view of AI:

  1. I can't control what anyone else does with AI, especially the corporations that operate LLMs, but I can control how I use it in my own work.
  2. It is currently impossible to avoid AI, and I do not foresee it becoming any easier. Objecting to it inherently will make me miserable because there isn't anything I can do about it.

Those of you who have followed my blog for a while know I am a journalist by trade (I have written extensively about it). This means, in my 25 years, I've written more than many people will write in their entire lives, both digitally and on paper.

I would not call the work I currently do "journalism." I am employed as a content producer β€” a catchall term that just means, literally, I create visuals and writing β€” and in my workplace, we use AI as part of the content production process. I push back where I can, but it is embedded in our content management system to an extent that it is impossible to avoid outright. The push to use AI in nearly every industry is palpable, as I'm sure many of you have experienced.

My AI use certainly informs my views, so I'll share the primary ways in which I use it.

I use AI for:

I do not use AI for:

I avoid consuming AI generated multimedia content, but it feels impossible to dodge at the moment.

What I've learned from using AI in my work

Many people already consider AI indispensable in their daily routines, which makes sense. Something that struck me when I first started exploring AI is how addicting it is. Especially in these particularly lonely times, it can feel like talking to a friend online. I find myself falling into that habit from time to time.

To that point, AI is great at mimicking human behavior. It knows how to sound like us and emulate our emotional responses. That also means it's great at making excuses.

LLMs don't like saying no. They very easily create false information out of thin air when it doesn't know what to do in response to a prompt rather than simply saying "I can't do that."

A very early use-case of AI I encountered was the embedded chatbox in Otter.ai, which I have paid hundreds of dollars to over the years for transcription services (Apple's AI-less voice memo transcription has grown leaps and bounds since, so I no longer pay for Otter).

Generally, I've found AI is pretty good at doing very specific tasks when given a very specific set of data to draw from. In the case of Otter's AI features, it drew only from the active transcript when you asked a question. As a journalist, it was really useful when writing breaking news.

If I remembered that a speaker at an event had something to say about roadwork, I could prompt, "What did politician Jane Doe say about local infrastructure?"

90% of the time, it worked perfectly and provided a list of quotes I could plop in my article. The other 10% wasn't so good. I realized pretty quickly that it was giving me completely fabricated quotes.

I spent a lot of time trying to wrap my head around why this could happen. What I eventually landed on was that β€” because the LLM doesn't like to say no β€” if the person speaking didn't address something I asked about, AI would treat the question as, "What would politician Jane Doe have said if she said something about local infrastructure?"

I started catching these potentially-catastrophic errors more frequently, and alongside it I started developing my own system for returning to quotes in the audio transcript after the interview (to those of you who had to learn shorthand to keep track of interviews and write down quotes: I am in awe of you), so my need to use Otter's tool eventually dissipated.

Other journalists didn't quite catch the AI errors in time.

Using AI at work has me feeling several different kinds of ways. On one hand, it can do a lot of my job. Like I said, it's great at emulating the way humans communicate. Learning to write is about pattern recognition, and robots love pattern recognition.

I do believe, however, that the work AI creates without the precursory research, fact seeking, interviews and local knowledge I provide is pretty bad.

My beliefs surrounding AI

Keeping in mind my previously stated worldview, I have a handful of opinions I've developed about AI over the last few years.

1. Corporations are taking advantage of consumer guilt.

I grew up in an area that was directly impacted by the Deepwater Horizon oil spill. In the aftermath of this scourge upon the Gulf of Mexico environment and its surrounding ecosystems, BP shifted its public relations strategies to focus mostly on assuaging the public's fury and somewhat take ownership.

Before this, the company framed itself as one of the most environmentally conscious oil companies in the world. Their marketing teams orchestrated a number of campaigns to push the narrative that BP cares about the environment.

In these ads, BP focused on what individual consumers can do to circumvent climate change maleffects. Many communications, environmental and sociology experts say this changed the United States' framework how we think about climate change. Namely, that we take on a disproportionate amount of environmental guilt.

There is much more to be said on this that I encourage anyone to read more about if you are interested. What I'm getting at is that mega corporations have built a reputation for pushing the guilt of their wrongdoings on the consumer.

I believe this is happening with AI, albeit more covertly. Of course, Google, Anthropic, OpenAI, etc. want you to use their products, but they'd rather us get pissy at each other rather than point our pitchforks at them.

[Aside #1 start] Data centers

Right now, a lot of the anger from the average American is toward the companies that are erecting data centers in predominantly low-income, low-education and low-employment areas (article gift link valid until Aug. 26).

The issue, as I see it, is not that the data centers exist in the first place. Pretty much all of our internet use relies on data centers, which have been around for decades. Rather, the issues are that:

  1. The data centers are by-and-large only being proposed in poor neighborhoods.
  2. The AI data centers absolutely gobble up electricity.

From the above Politifact article I linked:

Xu said a shopping website may require relatively little computer power for each request it gets, but training a large AI model can involve thousands of specialized processors operating together for an extended period. He also said that when large numbers of people send requests to an AI system and expect immediate responses, it requires substantial computing power.

So far, the political response to data center protests has mostly been passing moratoriums on their construction. I wish that stopping to carefully evaluate something unfamiliar was a more common approach, but I'll take what I can get. This is a perfect example of one of my worldviews in action: The most important work happens in one's own backyard. In nearly every case of these pauses on data center development that were passed by state or local lawmakers, it is a direct response to communities coming together against something that will likely make their lives worse without intervention.

I don't think it's reasonable to expect that these moratoriums will result in data centers not being built point-blank. My hope, instead, is that people smarter than I β€” and the politicians who are making these big decisions β€” come up with some way to reduce the environmental impact of future AI data centers.

[Aside #1 end]

2. Disclosing AI usage (or lack thereof) isn't really helping.

Digital creators on Instagram, Facebook and the like have been in shambles over the last several days since Instagram started showing that their content was created by AI because they used Canva or similar editing tools.

Many of them argue it isn't fair because they didn't know the tools were AI-dependent β€” to which I'd say we need to keep in mind as we teach folks what AI does and how it works. The cause of the "created by AI" tag seems to mostly be background removing tools. If they created all of the content in the graphic and just used AI to remove a background or change coloring, does that mean they didn't create the work?

The same goes for academics that are voicing concern over LLMs adding watermarks to text created with AI. For the writers using AI tools to edit their work, does the AI intervention mean that the human behind it lays less claim to it?

Conversely, some people are adding "created by a human" tags to their work. I started thinking about this after reading exomonstruo's blog post this week.

I honestly don't see the point in the designation. First off, if a reader gained something from a post, I don't think it's a big deal whether or not AI wrote it in part or in full. I have a decent time figuring out what text was entirely written by AI, but I'll admit I'm not perfect at it. Historically, I don't tend to find much value in purely AI-written texts.

Second, especially in a blog context, I think if you didn't use AI, your writing stands on its own merit. Most of the blogs I read on here seem personal and have their own unique style. The "I didn't use AI" tag reads like virtue signaling to me, which isn't inherently a bad thing, but I think it is pointless.

I believe that getting in the habit of believing everything is written by AI unless otherwise stated is harmful for creator communities and for the psyche. It further perpetuates an environment of distrust.

Two quick caveats:

  1. I think the disclosure tags can be useful in an educational context. When you see these β€” and you trust the writer β€” use it as training material to remind yourself what human writing looks and feels like.
  2. I see the value in using them to find community with like-minded values, but I think there are other ways to do this. Bear Blog's existence in and of itself is a great example of this. Folks on here aren't creating easily-skimmable content designed for an endless scroll; it is mostly thoughtful, slow writing that exposes the bloggers' thoughts, feelings and judgements.

3. Fear of being perceived to be using AI is making artistic, literary and academic content worse.

When reading fanfiction in the last few months, I have seen a tag appearing in most of the works.

Something to the effect of:

NO AI

or

AUTHOR DID NOT USE AI

I understand this compulsion with the climate surrounding AI use right now. Left and right, fanfic writers are being accused of using AI in their work. Some authors have shared their writing stats to prove they actually spent time writing without AI (This seems like a very cool tool... I loooove a statistic).

I prefer fanfic authors β€” or authors in general β€” rely on AI less and feel more confidence in their own written word. I think a lot of AI use in creative fields comes not from laziness but from insecurity in one's own skills. But do I think we should be spending our time bitching at fanfic writers we suspect of using AI? Not particularly.

What really stuck with me was coming across the guidelines for a fanfiction-writing competition. One of the rules was no AI, which is perfectly fine, but one of the ways in which the judges would "detect AI" was something along the lines of:

If there are minimal SPaG (spelling, punctuation and grammar) errors, this is an indication of possible AI usage.

I'm sorry... what?

I was among the final students to go through undergraduate without the help of generative AI β€” it existed in its current form for a little bit before I graduated, but it had not reached the point it has now where it can fully draft work that reads human. In my journalism program, a "SPaG" error was reason to dock a letter grade from an assignment. It was harsh, but it is true to the industry. Typos happen, but they should be rare. If a spelling error makes it to print, it is embarrassing.

Now, on LinkedIn, I'm seeing some people say they will purposefully inject typos in their work in order to prove they didn't use AI. Respectfully, you couldn't pay me to do that.

I am guilty of removing some AI "tells" from my writing. The em-dash has made a reluctant comeback in my personal style guide, but it was absent for a while in my professional work because of how widely it was perceived to be a sign of AI-generated text. Now, I'm seeing that colons are the newest AI giveaway. God forbid we use lists in our writing.

[Aside #2 start] How AI learned to write

Part of where I get my feelings regarding AI comes from my profession. Before I had even heard the terms "AI" or "LLM," my articles, which are publicly accessible, were mass-scraped from the internet to teach AI how to write.

Of course, it wasn't just me. Journalists, academics, authors, bloggers and influencers fell victim to this. There isn't anything we could do about it. We will never be compensated for the use of our work. I don't think it's fair or right, but it's the reality.

I'm not speaking anything revolutionary or new into existence when I say that all of these "AI tells" exist because it is how writers write. The robots didn't make up effective communication tools out of thin air.

I know that AI tells are intended to be a single indication that someone used AI to create work they are passing off as their own. i.e. 4 pairs of em dashes in a 300 word work is a yellow flag rather than 1 em dash being a red flag. However, that's not how it's being treated in practice on social media.

[Aside #2 end]

Where I think we're going from here

I do not think the way we talk about AI is sustainable in the long term β€” both the skepticism and the support. One side is averse to using it entirely, and the other wants to see it transform every aspect of our lives.

The reality of where we'll end up, likely, is somewhere in between. We know AI isn't great at everything, and while it will continue to get better β€” the idea of which does harbor some fear within me β€”Β there are many things, both art and consumer products, that people will prefer to have a human touch to. Companies who are laying off employees in favor of replacing them with AI processes will hopefully realize that the products they're making are demonstrably worse β€” if they haven't already.

A nonprofit I volunteer with ran a fundraising event a few months ago, and I created the marketing materials. It was nothing fancy. I have some design experience, but it was mostly just arranging text on Canva β€” sans the AI tools, as far as I know β€” and using the existing brand colors. I was shocked by the number of people who approached me to say they loved my graphics, especially because they could tell it was made by a human. It made the event more personal.

As far as marketing goes, many folks show a distaste to anything that is blatantly AI. You know... those flyers. I don't refuse to patronize a business or attend an event because they use tacky AI flyers, but it does make me think the product they are advertising will be worse. I'm sure AI will continue to improve at design, but I don't think it will ever have a fundamental understanding of what design principles apply to certain use cases and how to make a person feel something.

I have hope that humanity will have a realization that there is value in retaining humanity. It will be β€” and already isΒ β€”Β a long, treacherous road to reach that, if we do. Human beings are remarkably adept at adapting to new environments, and I think we'll prevail in spite of the slop.

πŸ₯œπŸ“Š

__

Let's talk! Shoot me a comment or start a conversation with me by emailing hi@nutgrafs.com.