A-Infos (en) Australia, ACF: Picket Line - The AI gamble (ca, de, fr, it, pt, tr)[machine translation]

Capitalists have turned to AI to keep profits flowing, whatever the 
cost. But don't blame the computers. ---- Everywhere you look these days 
you see it. Fake photos and videos. Fake art. Fake music. Fake news 
stories. ---- Whether we like it or not, AI has become a part of daily 
life. Type any question into Google and the first result will be an 
authoritative sounding (but quite possibly wrong) AI-generated response. 
Students hand in AI-generated essays that are then marked by AI-grading 
software. Companies use it to generate entire advertising campaigns for 
products hyped-up as 'powered by AI'. Walk down the street and there's a 
good chance you'll see an AI poster. Open social media, and good luck 
finding anything else.

'Artificial Intelligence' is a bit of a misnomer. We haven't found a way 
to make machines 'think' in the same way that we do. AI models have been 
trained to harvest (that is, steal) massive amounts of data. They 
compile words, images, video and sound, and, based on that training, 
predict the most likely assemblage of words or pixels when faced with a 
prompt. Ask for a picture of a cat and you'll get a prediction of what a 
cat probably looks like, based on millions of other images.

The AI companies can nudge models to favour certain kinds of responses, 
but the general tendency is toward a generic, uncanny 'average' - and 
with plenty of room for mistakes. The result is that the first big 
impact made by AI has been to make the world uglier and to make 
information less reliable.

But AI isn't just generating bad art. Today, it's at the centre of the 
global capitalist economy.

Nearly $3 trillion has been invested into the technology worldwide, 
including $300 million in Australia alone. Markets across the world are 
increasingly being propped up by just a handful of AI-industry stocks.

So how did we get here? Is the AI revolution all it's cracked up to be? 
Are the robots coming for all our jobs (or maybe even to take over)?

Falling profits...
To understand the hysterical mania around AI, you first have to 
understand the position capitalism was in when the technology arrived.

For capitalism to survive, the system needs to be constantly expanding. 
Investors need to be sure that production will lead to profits, and that 
they will be able to out-compete rivals in the market.

One way to beat the competition is by boosting productivity and cutting 
labour costs. The bosses are constantly on the hunt for new technologies 
and forms of automation that can give them an edge.

But the advantage is only ever temporary. When competitors jump on 
board, everyone is back on the same playing field. Companies can produce 
more, quicker, and with less labour. But over time this tends to make 
the value of the product fall.

This has been a major crisis for capitalism since the 1970s, when 
developed capitalist countries felt the pinch of falling rates of return 
on investment in industries like manufacturing.

The ruling class tried to solve the problem in a few ways. They 
privatised public services, turned the financial economy into a casino 
for bankers and stock-traders, and they sent huge swathes of their 
industrial economies overseas, so that unions couldn't win as big of a 
share of the shrinking pie.

This worked for a while. Until it didn't. The financial system built to 
make up for falling profit-rates crashed in 2008. Since then, 
capitalists have struggled to decide where best to park their piles of 
cash. In tech start-ups? Real estate? Ponzi-schemes, like crypto and NFTs?

Actually making stuff is of less interest. The ruling class is obsessed 
with 'stagnant productivity' - the fact that they haven't been able to 
squeeze more profit out of workers, particularly in industries like 
care-work and hospitality, where automation and out-sourcing is trickier.

... Rising bubbles
This is why the bosses are so excited by AI. Capitalists are staking 
everything on the hope that it will deliver a massive productivity boost 
across the economy and allow companies to throw millions of workers off 
the payroll.

The AI-boom, then, is a multi-trillion-dollar roll of the dice. And with 
all that money sloshing around, everyone wants to get in on the action. 
The result has been to transform the AI gamble into what is almost 
certainly a massive bubble.

Apart from the companies making the hardware for data centres, none of 
these businesses are actually profitable, or delivering on their 
promised potential. The computer-power needed to train and operate AI is 
insanely expensive. The sums dwarf the amount of revenue that these 
companies are bringing in from users.

AI stock prices are based on hope, hype, and some sleazy accounting 
tricks. Take the example of 'circular financing': Microsoft buys 
billions of dollars worth of hardware from a company like Nvidia - 
Nvidia uses that money to buy a stake in OpenAI (ChatGPT) - and OpenAI 
funnels that money back to Microsoft, for access to their data centres. 
A bag of money switches a few hands, expenses magically become revenue, 
and the stock market celebrates.

Some of the hype takes the form of fearmongering. Senior AI developers 
have made wild claims that their software is so powerful and 
revolutionary it could become self-aware and kill us all. The solution? 
To give them more power and money, obviously.

The trouble is, it's not just wealthy investors who are at risk if this 
whole thing goes belly up. The major tech companies have already issued 
around $200 billion in corporate bonds ('IOUs') to help fund AI 
infrastructure. That debt is being held by banks, governments, and even 
retirement funds.

So workers lose either way. If AI delivers on the promise fueling all 
this investment, millions of people will be thrown out of work. If it 
doesn't, the bubble bursts and the economy tanks - taking superannuation 
with it.

Burn the data centres?
Despite not quite living up to expectations just yet, the ruling class 
have still managed to find some uses for AI.

The creative industries are driving out real artists and replacing them 
with AI-generated slop. Gig-work companies like Uber are using 
AI-powered algorithms to surveil workers and customers, testing for 
signs of desperation so they can raise prices and low-ball wages.

States have embraced AI as a tool of repression too. In Palestine, the 
IDF uses AI targeting systems to carry out it's genocide. In the U.S., 
ICE uses it to generate possible targets for deportation.

Behind this Orwellian nightmare is the physical infrastructure that 
makes it all possible: data centres.

Data centres are essentially big warehouses filled with computers, 
running 24/7. Capitalist governments around the world are racing to get 
this infrastructure up and running so they can get their hands on some 
of the spoils of the AI boom.

There are some workers that are keen too. Construction workers and 
sparkies lucky enough to get jobs on these projects will be glad to take 
contracts on union-run sites with good EBAs.

Others aren't so happy. What will the extreme temperatures generated by 
the facilities do to surrounding soil, or the local water supply? And 
what good are a few union jobs if AI ends up putting the rest of us out 
of work?

Some of these complaints don't hold up as well as some on the left seem 
to think. For now, the water usage seems manageable outside of 
drought-ridden areas, and the 'heat islands' created by centres don't 
reach very far. As for noise, roads are worse offenders, and far more 
land is wasted on luxury golf courses.

The real problem is the sheer amount of electricity these things chew 
through, and the rate at which the demand for data centres is growing.

As we prepare this issue of Picket Line, the AI company Anthropic 
(Claude) has signed a deal to build Australia's largest data centre in 
Western Downs, Queensland. That one facility is projected to use up to 
three times more power than Brisbane's 1.4 million residents combined. 
That adds up to an extra 6.7 million tonnes in emissions a year.

As long as we're generating electricity with fossil fuel inputs, data 
centres will be big polluters. And as long as we live under capitalism, 
there will never be a limit to the amount of energy and raw materials 
that the system demands.

The future of communist robots
This has rightly led to a lot of fear and anger. Some have gone as far 
as saying that they want data centres and AI companies shut down entirely.

Slowing down the rate of data centre construction is probably a good 
idea. But we shouldn't throw the baby out with the bath water. As with 
prior forms of automation, the real issue is who controls this 
technology, not the technology itself.

Under the control of capitalists and politicians, AI will be used to put 
people out of work and wage war. Flooding the internet with ugly, fake 
garbage, and removing human oversight for crucial tasks is a small price 
to pay for profit.

But things could be different.

When the port logistics company DP World announced plans to replace 
workers with AI-assisted cranes and driverless vehicles, dock workers in 
the Maritime Union of Australia (MUA) responded by calling for a 28-hour 
work week, with no reduction in jobs or pay.

At the heart of the MUA workers' demand is the principle of workers' 
control - the logical conclusion of which is anarchist communism.

Under workers' control, AI could be harnessed in a totally different 
way. We could supervise these models to make life easier. We could take 
advantage of the genuine strides made by AI models in programming, 
mathematics, and medical science to meet everyone's needs. And it could 
all be powered by renewables at a sustainable scale.

While we wait to see if the bubble will burst, the capitalists are 
building a world where robots make the art, and humans do all the work. 
Our goal should be to turn it the other way around.

https://ancomfed.org/2026/10/the-ai-gamble/
_________________________________________
A - I N F O S  N E W S  S E R V I C E
By, For, and About Anarchists
Send news reports to A-infos-en mailing list
A-infos-en@ainfos.ca
Subscribe/Unsubscribe https://ainfos.ca/mailman/listinfo/a-infos-en
Archive: http://ainfos.ca/en