“The purpose of a system is what it does”
– Stafford Beer

“If privacy is outlawed, only outlaws will have privacy.”
– Phil Zimmerman

The proximate motivation of AI driven companies like Claude, Open AI, and Google is disruption: Silicon Valley slang for bankrupting incumbents by creating software platforms and skirting the law. Classically, this has played out as the replacement of secure often unionised employment and small business ownership, with precarious pseudo-employment. The process has helped accelerate financialisation, the mechanism by which all aspects of social life from housing to friendship are commodified. For the asset management class, financialisation ensures the growth of their wealth outstrips both the rate of economic grow and labour’s capacity to negotiate.

Facebook (now ‘Meta’) ‘monetises’ disconnecting people, by charging users of their services to share what they’ve written, created or organised. AirBnB worsens rental crises in cities where it operates. Uber substitutes well compensated sole traders with underpaid contractors, while increasing the cost of travel. Embrace, extend, extinguish.

It’s already clear that AI is well on the way to creating mass unemployment, initially in creative industries, followed by law, health care, programming and ultimately (as robotics develops) manual labour. Deep learning systems – fed on government funded scientific research and centuries of individual creative writing and art, financialise the space of ideas and creativity. Just as social media has financialised attention, romance and human connection. They are likely to fuel the greatest wealth transfer in human history.

Kevin Kelly coined the concept of the ‘technium‘, to describe the interconnected information-technological machine of modern life. He hypothesised this leviathan has its own motives – primarily to grow – cancer like, in interconnection, complexity and ubiquity. Today numerous online quasi-religious movements have accepted and developed Kelly’s teleology. As well as darker visions like that advocated by cybernetics researcher Nick Land. Most recently the Effective Accelerationists (E/ACC), led by charismatic quantum computer scientist Guillaume Verdon. E/ACC is championed by ‘capital allocators’ as notable as Andreessen Horowitz cofounder Marc Andreessen, and Open AI CEO Sam Altman.

The name of the movement parodies Effective Altrusim (E/A), the libertarian donation philosophy popularised by crypto-billionaire fraudster Sam Bankman-Fried. E/A was popularised by the ‘Less Wrong‘ community, established by AI doom monger Eliezer Yudkowsky. Yudowsky argues ‘poorly aligned’ artificial general intelligences will eliminate humanity. He suggests AI development be paused and regulated like nuclear weapons, until ‘AI safety’ is better understood. Verdon and his fellow accellerationists by contrast believe AI is a boon to humanity, but if it does happen to kill us all they ‘for one welcome‘ our robot successors. A view of the post-human future rivaling that of literal nazi satanists for its bleak implications.

Back on planet earth, the proliferation of AI seems likely to lead to an Orwellian censorship regime. Already image generators running on a mid-priced home gaming computer can create photo realistic pornography. As I predicted last year, budding porntopreneurs have deluged user generated sex work sites like Onlyfans with AI generated imagery. Any images can be created freely using these models, including illegal pornography and CSAM. This should be unsurprising, as this material was part of the training set used to teach these models. It’s a common misconception that advanced AI cannot be run on personal computers. While it remains wildly expensive to train such models, running them is often trivial. LLMs run locally on mid priced laptops, requiring no technical knowledge, can instruct users in the construction of a bomb, or how best to hide a body. They will within a few months be able to let you generate your own Marvel or Disney movie. Right now, all the pieces are in place to run interactive robocalls at scale, featuring the voices of your nearest and dearest – perhaps developing a relationship over days or weeks, before they request a compromising nude or money transfer. It’s only a matter of time until confirming the identity not only of a piece of news footage, but the other end of your Skype call, will be impossible. The same technology can be used to call your bank, employer or customers on your behalf.

Imagine the pressure being applied on legislators from every corporation whose business model is based around ‘intellectual property‘. Try to picture the terror every politician must have of election-eve videos depicting Steele Dossier like antics. The anger they will develop towards technologies that push young girls to suicide. The moral panics these technologies are likely to induce – as our capacity to fake imagery, video, voice and conversation accelerates will allow for only one solution.

Cory Doctorow predicted this a decade ago, in The Coming War on General Purpose Computing. The only way to prevent the creation of such materials is to embed AI snitch overseers and censors in every piece of computing capable of training or even running an LLM. Censoring the internet only prevents the sharing of illegal or problematic material. Censoring the computer prevents the idea being articulated in the first place.

Of course such spy chips will have the side effect of making private communication impossible, political organisation difficult, and the internet a kitten-soft non-threatening sphere in which piracy is impossible and creative distribution allowed at the whim of oligarchs. Its the end game of digital control.

Faced with this outcome the anarcho-utopianism of the E/ACC tribe seems tempting. Better a million LLMs chaotically dissolving societies capacity to inhabit a shared reality than a disneyfied corporate Brave New World?

Pick your dystopia.

_{All imagery in this article is AI generated.}