AI Slop is Eating the World

The notification arrived while I was trying to write an email: "Let AI help you craft the perfect message!" I dismissed it, only to have my photo editor suggest AI-generated backgrounds for my latest shots. My browser helpfully offered AI-summarized versions of articles I hadn't even clicked. In the span of five minutes, three different artificial intelligences had volunteered to do my thinking for me — all of them unwanted, all of them mediocre, all of them aggressively confident in their abilities.

This is how the AI invasion actually happens: not with superintelligent machines taking over our power grids, but with a thousand tiny digital assistants slowly eroding our ability to think, create, and communicate on our own terms.

The Age of AI

I firmly believe that in the future, even in a couple of years, an overwhelming majority of what you consume will be AI generated. This isn’t inherently bad, because AI can do a fine job of creating valuable things that people want. But today’s reality falls far short of that promise.

I think that everyone believes this because when we talk about what AI is capable of we mean one that’s better than what we have today. We pluck a frontier model from 2030 that has all the kinks worked out: it tells the truth, it has style, accuracy, and it operates at speed.

Yet it remains that there is a large delta between what we have today and what we see in the future, and it’s not solely the business of making the models better. We have to update our priors and develop a new relationship with the technology.

You’re not alone if this changing of the guard brings you distress. The human experience of consuming media finds comfort in knowing there’s another human on the other side of the exchange. Consuming media is an experience that operates within a tradition of human storytelling that we’ve been hardwired to feel. The places where AI has began to interface in our every day life are things that have long sought to be automated, but not always by us. Decision and logic trees plumb invisible paths underneath customer service chats and most humans know if you have something with more emotionality you’ll need to call and summon a real human who can apply a more human touch – empathy or a sense of humanity – to your request. One can imagine that the company uses this robotic chat experience to shuffle off customer support overhead and to wash out any human intervention that bends the rules in favor of customers. There are many cases like this, that appear to serve corporate entities and not the end user. It’s difficult to feel optimistic based on these priors.

Consider YouTube's latest initiative: a conversational AI tool for summarizing videos, suggesting related content, and answering viewer questions. Before you lies a forked path, on one side a human creation, on the other, a robot interpreting a human creation. This represents more than a feature addition - it's a fundamental shift in context. Users who come to watch videos are now nudged toward texting with a robot about the content instead. Like fast food promising convenience while delivering empty calories, AI integration sacrifices quality and connection (humanness?) for quick implementation. And as we've learned to be wary of processed foods, users are developing a healthy skepticism toward processed thinking and cold metal logic.

Humans are complex creatures capable of dizzying paradox. We readily welcome simple automations and help with verifiable logic, but eschew more human matters. The rise of AI therapists seems to cut this argument off at the knees but something more subtle can be seen. Some things, the crowd has determined, are too personal, too human for other people.

This isn't a simple matter of AI being "good enough" or not - it's about understanding where and when we want human connection versus algorithmic efficiency.

But there's a vast middle ground where things get interesting - tasks that require both efficiency and human understanding. This is where the real opportunities and challenges of AI integration lie.

Death By A Thousand Bots

I make a conscious effort to kick tires and try to build a more-than-surface level understanding of the products I use and it’s not clear that any of this new functionality is good for me. When I want to use an AI tool, I go to one of the major ones. These standalone half-baked integrations pushing up through the biome of apps on my phone only garner frustration.

It’s even working its way into scientific journals — the last great bastion of human exploration and advancement — and when it’s not being used to write, it’s used to manipulate peer review or to generate shoddy images from data, like rats with really, really big genitalia. Nature reports, “At major computer-science publication venues, up to 17% of the peer reviews are now written by artificial intelligence.” While science needn't be exclusively human-created, accuracy remains non-negotiable.

Abstracting a little, AI isn’t just taking over content. It’s working its way all the way down to the metal. Smartphones are becoming AI first devices. A cabal of trillion dollar companies believe you’re anxiously shuffling at the checkout line for the device with the best AI features and specially tuned chips. Only I don’t know anyone who actually is. The people I talk to fall somewhere between apathetic and highly suspect of any benefits AI integration might bring to their use case. One exception to this is group of software engineers that I keep as friends. But there is a bitter irony to our use and that is that every time we commit AI magic to our code base it’s accompanied by an uneasy feeling of deprecation and a realization that this professional road might have an end before we’re ready to retire. I can only imagine how artists and photographers must feel, having their profession become the quintessential AI proving grounds for the public.

Winner Takes All

I do think that hardware or operating system level AI integration is preferred to a slurry of bespoke interfaces. The effect of having 50 different apps with 50 different AI functions is that I don’t want to use any of it. The rapid pace of AI development means today's cutting-edge solution is tomorrow's abandoned platform. It feels like these one-off solutions will be quickly outmoded and eventually sunset, marooning any dependent users they’ve managed to win over in the process. It would be much better to have one AI tool to rule them all that can interact with my apps or websites that I can invest my effort and time into honing in. However, as we’ve seen with the internet at large, aggregation and consolidation must be avoided at all costs. The future needs to be open-source, and we need to have devices capable of using it. Unfortunately, the path to such a system remains unclear.

Moreover, our current interaction model with AI, primarily through chat prompts, feels like a temporary solution. Text-based interaction is inherently slow and cumbersome, suggesting that a revolutionary AI interface — one that truly transforms how we work and create — has yet to emerge. I can’t shake this lingering feeling that there must be an optimal user experience that doesn’t feel as much like a retrofit for what we have today.

Much of this piece has stuck to the confines of computing devices, but another AI frontier is robotics and things that interface with physicality in the real world. Those feel further off and feel largely predicated on what we can achieve, fundamentally, in the digital realm. It could be that I'm wrong about this and that considering an intelligence first approach and applying it to movement isn't the approach that will ultimately win in that space, but that seems to be how we're going about it today.

Fundamentals

As I finish writing this the news cycle has shifted to apathy. OpenAI and others are having difficulty achieving smarter models. Use has stopped growing. Apple has yet to ship their fashionably late addition to the AI frenzy.

Maybe it’s safe to say the honeymoon period is over. Or maybe it’s just a seasonal lull as the foundational grid is updated beneath our feet for a blistering set of ‘25 releases.

The fundamental challenge isn't just AI's inconsistency - it's our relationship with its mercurial nature. Each interaction feels like reaching into a grab bag: sometimes you pull out something brilliant that genuinely advances your work, other times you get plausible-sounding nonsense that sets you back hours. We're caught between refusing its help entirely and surrendering to its seductive promise of effortless productivity, never quite knowing which version we'll encounter.

As the initial AI frenzy settles into a more measured reality, we have an opportunity to shape its integration thoughtfully. We need to build digital spaces that amplify rather than automate our humanity - spaces where AI serves as a collaborator rather than a replacement, where technology enhances our creativity instead of supplanting it, and where we can trust our own judgment about when to engage and when to step back.

In the end, perhaps the real threat isn't AI itself, but our tendency to accept mediocrity for the sake of convenience. The challenge ahead isn't technological - it's cultural. Can we resist the allure of AI-generated adequacy in favor of human excellence? Can we harness these tools while maintaining our capacity for original thought, genuine connection, and authentic creation?

The answer won't come from an AI model, no matter how advanced. It will come from us, from our choices about how we integrate these technologies into our lives, and from our commitment to preserving what makes us uniquely human in an increasingly automated world. Our future depends not on the sophistication of our AI tools, but on our wisdom in using them - and our courage to sometimes choose not to use them at all.