The (Two) Big Anti AI Rant(s)
I’ve been posting some long-ish writing pieces on my Facebook wall recently but Meta is evil and the algorithmic internet is stupid so I’ve decided to resurrect the blog here on my actual website. Remember when the internet was more than just 5 walled gardens owned by technofascist billionaires? If you’re very young you may have missed that incarnation of it but it was better (I’m gonna write a piece on that soon).
These aren’t really comedic but I’ve had some positive feedback from readers so this felt like a decent place to post them. Substack supports right wing evil fucks so I decided not to put these up there for the moment (no shade if you write on there there’s no ethical consumption under capitalism ofc).
If you follow me on socials you may have already seen one or both of these posted rants, in which case there is nothing really new in the below text – I’ve done some minor edits for formatting and clarity but the essential text of both is the same.
I also sent this out in the email blast so, you know, you may have seen it there as well.
Using Generative AI is Bad But It Doesn’t Mean You Are A Bad Person
I’ve tried to say as much below but if you currently use AI for stuff I would ask you to consider stopping, and I also want to make clear that just because you use AI doesn’t mean I’m calling you a bad person, but I would ask you to please consider the impact you are making on the world, as AI is making this civilisation we share a worse place to live.
Okay here we go:
TERMINOLOGY
First thing’s first: “AI” as the term has been used these last few years is really, really inaccurate.
Large Language Models (LLMs) like ChatGPT are not any kind of ‘intelligence’, as they are not designed to have any understanding of the words they are guessing should complete their text chains.
“AI” sounds a lot better than “word-guessing machine” and most of the “AI Revolution” is driven by the marketing arm of the mega corps and billionaires who stand to profit by it (though the whole industry is burning money at an insane rate – for no real benefit to anyone, even if you currently THINK you are gaining benefit). The economics most clearly track to the subprime mortgage crisis, and crypto boom. When we say ‘profit’ here its essentially profit as determined by grifting the public, the stock market, investors, governments and megacorps.
When it comes to image generators, this again is the result of an averaging-machine. There is no sentience (nor path to sentience) involved in the process.
For clarity, I will refer to them both as “AI”, or “Generative AI”, even though the term is a misnomer.
The following is from two posts I made to social media – one on February 9, and one on May 7. Both have been slightly edited for clarity.
THE ANTI-AI-ART POST
Originally posted Feb 9 2026, minor edits yesterday
This was posted in response to the trend of posting “AI Art” interpretations of oneself.
Let's do a trend where we all pay artists to draw something instead.
LLMs and GenAI-driven image generators waste massive amounts of electricity, drive the recession, waste fresh water near communities, drive up the price of ram and graphic processors, steal from artists, and degrade the internet & the utility of websites.
They do this so that mega corps and multi billionaires can chase the next hypergrowth market and cut employment opportunities for everyone. They do not and can not by definition create a better output than a human can create. They are averaging and probability guessing machines.
What they CAN do is be owned by mega corps and multi billionaires (although, ridiculously, the actual costs of running them will never, ever, make financial sense to the value of their outputs - they are the new crypto and are currently running at a loss of many billions of dollars)
Anytime I see someone post AI-created slop I unfollow them, and I encourage you to do the same
The big problem of the internet is not A Lack Of Content. There is more ‘content’ and entertainment already in existence for you to be entertained for multiple natural lives.
The problem is saturation of low-effort and low-value slop that makes discovery of content difficult. There’s loads of good stuff out there that you are not seeing because the ocean of the internet is polluted with slop.
Put some fucking effort in. Express yourself, rather than offloading your creativity and cognitive load to an ocean-boiling machine.
I'm consistently broke but I paid an artist $350 to draw the animals (and me onstage) for my comedy showcase in Perth, and I'd do it again. I paid the same artist to draw the rabbits for Bun Runner / The Smoking Bun years ago.
If you use ChatGPT and generative AI-driven image generators you are actively destroying the internet
I'm not saying you're a bad person, but you are doing a bad thing. If you are an artist using these things I honestly find it almost impossible to understand your values.
If you use ChatGPT to answer questions and "do therapy" you are actively destroying your own ability to think creatively and critically
The way LLMs are constructed is not to "read the internet and give you the answer". They are constructed to guess the next most likely word for you, the reader, to accept as 'probably correct'. That means it will flatter you, and lie to you, as long as you are likely to believe it. It understands nothing, and limits your ability to understand. It is a sycophancy machine and we all like to be sucked up to.
Even if you use the services for free, you are helping to prop up a destructive and parasitic system.
Every free use of Generative AI goes into their stats to waste 10s of billions of dollars more in investment capital. The global recession is in large part driven by this HUGE drain of humanity's resources into the word/image guessing/stealing machines.
If you pay them money you are collaborating in the enshittification of the internet in a very active way
LLMs are not accurate. They are not better than learning a skill or researching a question yourself. They function in a way that they make the sources they pull from stop being viable in the mid to long term. They scrape websites without supporting websites. They steal from artists without compensating them.
Let’s not forget the old adage “if you pay nothing for a service, you are the product”
Google made their search worse deliberately to sell more ads. This is a matter of public record.
They, and Microsoft, are jamming generative AI into all their products because they overleveraged their companies into this wasteful, destructive bet on hypergrowth, the AI shit isn't better, and they realised okay we need to literally prevent people from having the choice to not use it (because it is worse), so they now they tell you it is unavoidable. They say “it is inevitable” because they want you to stop resisting. They are off-brand Thanos wannabes and they suck.
Divest from AI. Use other services.
Don't support and be complicit in wasteful, destructive slop.
When you see people using AI art know that they are saying: “we don't value art, and we don't value you.”
THE ANTI-LLM POST
Originally posted May 7 2026, minor edits today
I’m still mad about AI here's some more thoughts so I can get them out of my brain.
Generative AI Is An Opiate
And like all opiates it makes you feel good, while causing harm to you, and society, while enriching a bunch of criminals.
The reason we have laws regulating opiate consumption in society is because we human beings LOVE to feel good and we will pursue that over our own wellbeing – and if the good feeling is good enough we will do that until it kills us.
The Cost of AI Is Not Decreasing (It Is Increasing)
I see this one a lot. "Oh man it will be so cheap to run!" "Costs are coming down all the time!" "We just need to scale it up more!". Generative AI’s underlying architecture is not built in a way that the costs come down over time. It’s built for the opposite.
AI is getting more expensive to run over time, not less
Each consecutive LLM model is exponentially more expensive to train (& run) than previous models, for an incrementally improved functionality. The image generators (such as Sora, which OpenAI has now shuttered, in a move which meant they had to also cancel a 1 Billion Dollar Deal With Disney) are even more crazily expensive to run than the LLMs.
Every AI company has been setting money on fire and the current recession is in large part to the rush of venture capitalists to hurl their money on the blaze.
Generative AI "Productivity Benefits" Are A Lie
There are little (to none, to negative) "productivity" benefits to using AI. The "productivity" being sold is a lie the billionaires and megacorps are telling you because they need people to keep using it even though its not making work easier, because they need more capital investments into it, or their whole house of graphics cards will fall apart.
What I'm saying here is that at times using AI makes work take LONGER - this is not even accounting for the current subsidised running costs.
Other times it mostly makes it take the same time, or possibly very slightly less. And usually for a worse end-product than if you’d written the thing yourself from the beginning.
All reputable studies have shown the actual use-cases for AI are extremely limited and in most cases make the effort to work ratio worse. "Work" in this context being something usable to a professional standard out the other side.
Claude Code Is Bad Actually
You really do not want to use something that has been ‘Vibe Coded’
"Oh but Claude Code makes programmers redundant!". Even when Claude Code does successfully make something that "works", the code is insecure and buggy and verbose (meaning, there's way too much of it). Good code is simple, annotated well, and lean. Claude code is very wordy and way longer than it needs to be, convoluted, and has no useful annotation. This means it takes ages to run, has a bunch of bugs in it, a bunch of security holes, and is really difficult to debug and improve by an actual programmer, because its just a bunch of regurgitated code from github (fine to use) but copy/pasted with no understanding of how it works.
Vibe coding is just another Generative AI trap that can very quickly (for a lot of money that someone is paying, even if you are not) get you to something 70% usable, but then you are on your own to get it to 100%, and unless you already know how to do the thing then good luck coz that’s the hardest 30% of the work.
Why You Think Its Good (If You Think Its Good)
This is based on the Dunning-Kruger effect, which you've probably heard of. Know that everyone suffers from this, including smart people (sometimes they are MORE vulnerable to it because they "know they are smart").
My fiancee got a concussion a year ago and arguing with her about what medical precautions we should take due to her head injury was incredibly frustrating, because she remembered she's smarter than me - but she was using a damaged brain that was making her actively dumber in that moment.
Famed annoying smarty-pants Richard Dawkins demonstrated this a couple weeks ago, where he talked to a chatbot for 10 minutes then went "ITS ALIVE!" because he lacked expertise enough to understand that a word-guessing machine, talking to a person built by evolution to find humanity in everything, would seem like humanity. This is a natural extension of Pareidolia (or the ability to find faces in anything). It’s hardwired by our evolution.
The whole myth of vitamin C being "good for a cold" or to fight the flu comes from a super smart guy that thought because he was smart in chemistry he would also be smart in biology.
Linus Pauling got a Nobel Prize in chemistry and it made him so confident in himself he manufactured a completely wrong idea about vitamin C and infection (in the 1970s) that persists in the culture over 50 years later.
Vitamin C does not help to fight colds and flu btw
This is as accurate as 'vaccines causing autism', for similar reasons (extrapolation from a non-representative small sample, though in the case of Andrew Wakefield's work he probably falsified data in addition to using bad science).
Autism is genetic. Part of the reason for the increase in cases is diagnostic expansion & understanding, and part of that is once autistic people created the internet, more of us were able to find each other and have kids together (shoutout to Tim Berners-Lee and a bunch more autismiacs for making The Cool Internet before these mega billionaire fuckheads decided to railroad us all into 5 or so walled gardens that exist almost entirely to be a vector for advertising).
Nobody likes to hear they are bad at something, but in most cases where you think AI is helping you it is because you lack enough expertise to assess the AI's output correctly. Due to unavoidable hallucinations (integral to the technology) you can never trust AI output without having sufficient expertise to vet the output yourself. What this means is you can only reliably use AI to generate something you already know how to generate. In any other use case you are likely to let hallucinations through.
"AI" does not have any understanding of what it is saying. It is a word guessing machine. It cannot assess its own output. Only you can do that. That's why it is so nice to you. People like when someone is nice to them so they are more likely to accept the output as "valid" when it tells you that you are good / smart / correct / handsome / pretty / loved / lovable. It is a "Yes Man", and like all Yes Men it is therefore both basically worthless AND addictive.
The Real Cost Of AI Is Astronomical
AI is not worth it to run, even based on the subsidised costs they are currently charging you.
This is not even accounting for the environmental and power costs to train and run it. Currently that cost is being subsidised by billionaires and megacorps, but the crunch is starting and they are increasing the end-user costs. This is only the first step for cost increases - the actual training & running costs are fucking WILD and if they ever start trying to charge anything NEAR those you would never, ever, ever consider that a worthwhile expenditure of your own money.
They would very much like your dependence now, however, so you might feel there was no way to do certain tasks without AI
Dependency & Preventing Your Own Mastery
There is clear data that using LLMs and AI degrades your own cognitive abilities and reduces your ability to develop mastery. Work can be its own reward, because YOU GET TO GET BETTER AT STUFF. You cannot learn from AI. All you can learn is more and more dependence.
And ultimately all AI can do is make something that is the average of a bunch of other things. They have really successfully branded GenAI a totally wrong term. It is not “artificial intelligence” because it does not, can not, and can never, understand anything. That’s not what its built for.
You can though. You can understand stuff. Don’t cheat yourself out of that.
YOU can make something BETTER / DIFFERENT / INTERESTING because you have the ability to not only develop mastery but also to synthesise different ideas and inspirations together. Unfortunately developing mastery also means making stuff that sucks for awhile and that feels bad.
Opium feels good. But its bad FOR you.
Delete ChatGPT.