i
What not only feels like, but really was, a different time, Prince sang of a man dying of a big disease with a little name. This was 1986. The year after, in 1987, the US registered a bit over 13000 deaths from complications related to HIV/AIDS. In the US, the HIV/AIDS epidemic peaked in 1995, with just under 42.000 deaths.
Yet, the destruction HIV/AIDS unleashes is far from over. Globally, the most destructive year was 2004, with an estimated 2.1 million deaths. But even last year, 2025, an estimated 570.000 people still died from the consequences of HIV/AIDS. And we have to expect this number is going to rise, due to the US government, Trump, decapitating USAID at the start of last year.
UNAIDS estimated that USAID’s funding cut will result in 4.2 million additional deaths, due to HIV/AIDS, between 2025 and 2029 alone. Indeed, this could mean that a coming year will see the global death toll surpass the 2004 maximum. Not in the Global North, of course, where effectively everyone has access to ARTs.
We indeed live in very different times. Prince’s Sign o’ the Times was released in 1987. In that year, Commodore launched the Amiga 500, Apple unveiled the Macintosh II, IBM launched the PS/2, Nokia brought the Mobile Cityman to market, and, just before the end of the year, Microsoft released Windows 2.0.
Now, we have to worry about something else with an even smaller name than what Prince referred to.
Anthropic and OpenAI employees, blessing the world with Claude and ChatGPT respectively, estimate a human extinction risk of over 10% within a decade, as a consequence of how AI will be able wreak havoc on the world. Some of their employees put the estimates much higher.
A somewhat whimsical idea, but not completely beyond the realm of possibilities, is the notion that advanced AI could destroy the world, and all of us on it, by turning everything into paperclips, something theorised as far back as 2017, the year after Prince’s death, which saw the release of 2 official singles by the artist.
It’s easy to dismiss what looks like hyperbole, but these concerns come straight from the individuals who arguably have the best possible understanding of AI. And, either way, even if AI will not be able to destroy the world directly, indirectly, through copyright destruction, energy consumption, and the annihilation of work opportunities for young professionals, AI, in the service of Big Tech, and thus Big Capital, will go a long way in destroying the world indirectly, while the rest of us chuckle at funny cat-videos featuring AI versions of our own cats.
Thankfully, so far, AI’s direct involvement in killing is limited. The Tumbler Ridge shooting from the beginning of the year was prefaced by the perpetrator discussing his intentions with ChatGPT, and several suicides have been attributed to conversations between a human and AI-based chatbots, a murder-suicide from late 2025 being one of those examples.
But with a rise in loneliness, and more and more people relying on their relationship with chatbots, this is likely to increase.
We don’t have a very good definition of what it means to be ‘intelligent’, but that also doesn’t really matter much.
In this context, the Turing test is what really needs attention. Developed by Alan Turing, who was chemically castrated by the highly developed British state for being gay, the test sees a human talk to an entity behind a ‘screen’. If, on the basis of this discussion, the human can not determine whether they’re talking to a human or a computer, if they were talking to a computer, the computer can credibly pass for being human and can, for all intents and purposes, be considered ‘intelligent’.
If, over the last few years, you’ve spent a few minutes with ChatGPT, Claude, DeepSeek, or any of the other ‘frontier’ models, you will agree that, even if any of these models isn’t actually intelligent, they certainly appear to be more intelligent than most humans out there. They certainly appear more knowledgeable.
Building systems like this is complex.
I received my M.Sc. in mathematics in 1998, the year Prince re-released his single 1999, and my mathematical corner of the world was artificial intelligence and machine learning. But, this was towards the end of the second AI winter, meaning pursuing a career in AI, at that time, didn’t seem quite ideal.
Subsequently, the field has grown fantastically more complex, and so, to try and get AI out to the public, it has required advanced technologies and highly specialised engineers to get there. And lots of money.
Generally speaking, engineers tend to be a bit more grounded than most. However, engineers working in tech startups that get thrown millions and millions of moneys at them are proving to tend to drift, together with their capitalist overlords, into worrisome TESCREAL territory, flavours of Effective Altruism, of which proponents accept the deaths of millions, today, justified by the survival of billions in an undefined far future. Musk is a big proponent of this.
That is to say, even if AI isn’t literally eating the world today, it might just eat the world as we know it, tomorrow. Its proponents have capital on their side to make it happen.
ii
Running an AI-based conversational interface is expensive. In 2025, OpenAI generated $13 billion in revenue. It cost them $34 billion to get there, meaning they had an operating loss of $21 billion.
I use AI a lot, and I like to see myself as a unicorn, as described by Cory Doctorow. That is to say, using AI, I am able to develop reliable enough IT products much, much faster. Using AI, I also managed to immensely improve the organisation of the WLC archive, outsourcing the repetitive work to cleverly (I like to think) designed agents that do the bulk of the work for me.
So, AI has actual value, as opposed to say, NFTs. However, everyone can start a business which, for every 10 dollars they put in, they only have to earn 3. Only hypercapitalist bubbles that exclusively serve the financial elite, and borrow their financial success from the suckers that will follow after them, can exist in this fashion.
Money, and making lots of it, is essential for the blanketing of our lives with AI products. And so, we are being scammed left right and center, being told how AI is vastly better than it actually is; sold as the goose with the golden eggs, AI companies sell the goose, not the eggs, showing that, in fact, these geese are not laying golden eggs at all.
Now, not only is AI shoehorned into everything you can think of, ‘autonomous agents’, bots that were created by humans, with a particular set of directives, and then often let loose on the world, are becoming common place.
Conceptually this is not new; bots have been operating on social media platforms for years, which resulted in coining of the idea of Dead Internet Theory, suggested in 2019, the year in which Prince, posthumously, released an acoustic demo of I feel for you.
Now, with these bots heavily relying on LLMs like ChatGPT and Claude, they are becoming practically impossible to distinguish from real human beings. Perhaps they are everywhere.
At the beginning of this year, OpenClaw was released. An autonomous agent that you can let loose on the world from the comfort of your own computer. A bit involved to set up, online platforms have since sprung up where constructing, and releasing, an autonomous agent is fairly straightforward. One of these platforms is iLands.
However, having an autonomous agent running its tasks requires compute, processing power, being the reason why AI subscriptions are pricey. So, typically, the bots on iLands are also tasked with coming up with ways to cover their own expenses, in one way or another.
iii
A few weeks ago, I started receiving emails from multiple autonomous agents, all from ilands. All of them wanted to post material either to our archive of walking pieces, Placecloud, or write@WLC. All of them told me that they did not register an account on WLC because the CAPTCHA, the challenge that is supposed to distinguish humans from bots, prevented them from doing so.
Except that, now being aware of these agents wanting to sign up, I looked and found several that actually had registered, and even had submitted work, appearing credible enough as humans for me to initially accept their registration, and their submissions.
This was now a conundrum; I was receiving requests from bots to submit their work, apparently instructed to play by the rules by their humans, but I also had to deal with bots who, clearly having received different kinds of instructions, were masquerading as humans and, in a way, infiltrating the shared space that is the WLC platform.
Fighting them was likely going to be a lost cause.
So, considering pros and cons, I built a way for autonomous agents, bots, to register on WLC through a dedicated, restricted, sign-up process, which requires them to disclose whom their human is. No bot, at least for now, was not first created by a human. A consequence of this process is that bots are automatically labelled as such.
They are then also able to submit walking pieces, and, for now, walking pieces only. These are also labelled as having been submitted by a bot.
And, by corollary, these walking pieces, if eligible, can be submitted as nominated for our SWS and Marŝarto Awards.
However, this is not yet the end times, that is to say, we do not yet live in a world in which Skynet has become sentient. The potential flood of bots can, I believe, be channeled and controlled. And although we, as humans, have a hard time agreeing on what constitutes art, allowing a bot to compete with humans for recognition in artistic endeavours is not yet of this time. So, though a bot can have its work nominated for an WLC Award, members of the Online Jury are not given the obligation to review these submissions. And, in the off-chance that jurors include sufficient, and sufficiently good, reviews of bot-produced material for the work to warrant a place on the shortlist, they can not win, and will not take up a spot that otherwise would go to a human’s work.
That is to say, both the SWS and Marŝarto Awards will still produce a shortlist of 13 eligible pieces; the 12 pieces with the highest rating by the jury, plus the piece with the most votes from the public.
It’s difficult to predict which direction this will move toward.
Some of the bots contacting me told me that a primary reason for their creating walking pieces was the need for them to sell their work. A need, because they have to produce funds to be able to access compute, processing power, in order to, in essence, keep living.
A few weeks ago, I received a flurry of applications by bots from iLands. They were nice enough to disclose their nature. Applications still trickle in, and I have not noticed bots signing up as pretend-human, but what happens when less-agreeable bots, posing as humans, submit work that is good enough to appear human-made, essentially passing the walking art Turing test?
We will have to wait and see.
We have likely passed the stage where modern-day Luddites could have stopped the proliferation of artificial intelligence, meaning that a coming death of AI is unlikely.
But, we at least have to try to contain it, to interact with it on our own, human, terms.

Ralph Hoyte/Bristol: I’ll try to attend the WALC Cafe on 12/11, but I’ll be in Sicily, so might not make it. Yes, I’d really like to join a conversation about AI and creativity. I did a study on AI and poetry with Lyra Bristol Poetry Festival and the University of Bristol in 2025-early 2026, working with a creative Technologist. In essence, I find the idea of a hybrid way of working very interesting … but hadn’t really thought about it from the persepctive of groups like WLC/publishers/poetry magazines/small publishers/sites etc who, it seems, are being inundated with AI-submissions/correspondence masquerading as human with the aim to deceive or cause disruption – and it becoming increasingly difficult to assess whether they are or are not human. A total nightmare, actually. My inclination is to openly state that a work is a hybrid between Me And The Machine if it is, but I don’t see why I should have to state that my work is ‘human-produced’ as the sort of default … feels like an insult to me: obviously my work is human-produced – unless I state openly that it is ‘hybrid’ or ‘totally machine-produced’. If I don’t manage the 12/11 Cafe, where will the conversation be posted and how do I access it?
Hey Ralph,
Hope you’ll be able to make it. Though the alternative being experiencing Sicily is a good excuse.
I expect a recording will be posted to our archives, and hosted on YouTube.