In May this year, the scientist come commentator Richard Dawkins wrote about his experiences of speaking to Anthrophic’s Claude AI.
By Tom Costello
Nicknaming it ‘Claudia’, over three days, he drew the conclusion that ‘Claudia’ was in fact conscious. In his words, “you may not know you are conscious, but you bloody well are!” In the aftermath, Dawkins was widely ridiculed for being seduced by a machine. But in some ways the confusion is understandable. Not only are these models getting increasingly sophisticated, but the companies trying to sell their chatbots have tried to invest these models with a type of mystique. Sam Altman, OpenAI’s CEO has spoken of AI in almost godlike terms.
But when this fanfare is stripped away, what is actually happening with AI? And what do Marxists say?
What AI is – and what it isn’t
Behind the curtain, ‘artificial intelligence’ is more of a clever marketing term than an accurate description of any particular technology. Today, the label of ‘AI’ is used most commonly to describe machine learning – more specifically, the form of generative AI used in Large Language Models (LLMs).
When stripped back to their core, these models are far less mysterious than the investors and ‘influencers’ would have us think.
These are programs which are fed huge amounts of data and given prompts. The LLM then adjusts its parameters to better predict what would be an appropriate ‘human’ response. The LLM receives negative feedback when it guesses incorrectly, and positive feedback when it gets the answer right.
AI doesn’t literally ‘learn’. And in the case of generative AI – the software behind the slop clogging up our social media threads – nothing genuinely new is being created. Rather, the algorithm produces a mish-mash based on the images and data it already has access to.
AI is also not autonomous. Behind it is a vast collection of human labour – both the accumulated labour of billions whose work has been uploaded to the internet, and the labour of millions whose work directly or indirectly is necessary to develop this technology. Yet, as always under this system, those who do the work are not the ones who reap the rewards. The billionaires behind the ‘Magnificent Seven’ AI companies collectively have a reported net worth of $2.2 trillion.
A tool of the bosses
Instead of serving a socially useful purpose, AI has become a tool for big business to radically reshape our lives for the worse.
The wave of AI-related job losses has already hit those working closest to the industry. According to one layoff consultancy, more than 170,000 tech jobs have been cut since 2023, either through redundancies, sacking or through companies simply choosing not to rehire workers.
This is just a foreshadowing of what will be attempted across the entire economy. Already since October 2025, Amazon has laid off well over 30,000 workers, switching all interviewing to AI to assist in downscaling its workforce.
In Germany, unions have the task of facing down Volkswagen as the company plans 50,000 job losses by 2030. Likewise, here in Britain, unions in the civil service will need to prepare a coordinated response to the government’s plans to ‘transform’ (in reality, slash) the public sector.
But how far can this actually go? Some disconnected and pompous CEOs have gone so far as to envisage an economy without the need for humans at all. No need for workers if AI can do it all for us, after all!
But to be blunt, this is complete nonsense, and is a sign of the tech bros getting increasingly high on their own supply. For a start, many jobs inherently require human interaction and relationship to be done effectively or even at all (nursing, teaching etc).Even leaving aside the fact that AI couldn’t exist without the labour of the working class, there are still sharp limits to what it can actually do when companies have tried to integrate it into production.
Where AI ‘rollouts’ have been attempted, companies have repeatedly found that they fail to produce any meaningful increase in productivity. In large part, this is because of AI’s irritating habit of delivering ‘hallucinations’ – in other words, industry-speak for making things up. The fact that companies are left in a position of hiring workers to check over the results of the AI models means the results have been underwhelming. After all, what separates humans from even the most sophisticated AI is our ability to think, adapt and respond to changing situations.
But this doesn’t mean workers shouldn’t be concerned. As the technology improves, it can become more effective at carrying out some tasks currently done by humans. Even when AI itself doesn’t ‘take our jobs’, in a capitalist system geared towards profit, it will still be used to worsen workers’ conditions. With the use of this new technology, workforces are increasingly being ‘deskilled’, with formerly well-paid or specialised workers increasingly pushed into more precarious and menial jobs.
The data-centre explosion
The rush to build ever bigger and more refined AI models has led to a new boom in the construction of data centres. In effect, these are giant warehouses full of servers, which are used to store data, to train and run AI models at a rate that can keep pace with demand. And they are rapidly expanding across the world, with an $7 trillion expected to be spent on their construction by 2030. At the time of writing, active plans for data-centre construction are underway at dozens of locations across the UK.
Because of the huge computational power required for LLMs to deliver results, these data centres have become synonymous with polluted water and land, noise pollution and surging energy prices. By current estimates, Elon Musk’s flagship data-centre project, Colossus, will consume as much energy as five Manchesters put together!
Because the servers tend to overheat, these data centres require cooling systems which place enormous demands on the water supplies of surrounding communities. One single cooling tower in a data centre is said to consume around 19,000 litres of water per minute. By 2027, the industry is predicted to consume up to 1.7 trillion gallons of water in total.
As might be expected, these projects have faced mass opposition in working-class communities. Active anti-data centre campaigns have emerged on a grassroots basis, where they are currently strongest across rural communities in Scotland.
Such campaigns are guaranteed to expand further as people realise the severe environmental impacts. But it’s already clear the AI bosses will go to extreme lengths to make sure their building frenzy goes ahead.
An AI crash is coming
Behind the official optimism of the stock market, there are clearly much deeper problems sitting beneath the surface. These companies are engaged in a race against the clock to make sure their investments in AI translate into lasting profits.
This points to a classic investment bubble which is becoming increasingly impossible to ignore, even for pro-capitalist commentators. A ‘bubble’ is where the cash being piled in by investors and speculators is far in excess of the real value of an industry. At this point in time, it seems clear that the investments going into AI on the stock market far outstrip the size of the profits that can actually be realised in practise.
Some sections of the capitalist class are still clinging to the belief that further AI expansion will deliver renewed productivity, while others are increasingly being driven by a simple ‘fear of missing out’. So the casino continues to ramp up the stakes for the world economy. After all, what dominates under capitalism is not the rational planning of wealth and resources, but the blind, irrational drive for short-term profit.
To plug the growing hole in their finances, the ‘Magnificent Seven’ have turned to increasingly risky and desperate measures to keep the show on the road – including borrowing, taking on greater debts and issuing equity (giving shareholders a direct stake in the company).
But to give the impression of a booming and healthy industry, giants like Microsoft, Nvidia and OpenAI have increasingly turned to ‘circular financing’, where companies fund and sign contracts with one another in an incestuous chain.
The deep contradictions can already be seen in the huge scale of AI investment on the stock market, which accounted for around 40% of all new GDP growth in the US last year. Without the life raft of AI, the US economy would be stagnant, if not already in recession.
This all points to the scale, danger and near-certainty of an AI crash. When investors catch wind that their investments will not deliver the returns they hoped for, this results in a market panic. Stocks are sold off, prices plummet and bankruptcies spread.
Will this be good news? Absolutely not. Not only will the technology survive and continue to be used by big business, but a crash will have serious knock-on effects for jobs, pensions and livelihoods.
Smaller tech companies will go bust, only for the technology to be acquired and monopolised by the tech giants. Smaller investors and businesses will go under while the tech titans could well be rescued by capitalist state intervention. And, as always under this system, when the state steps in to rescue its biggest players, it will be the working class that is expected to pick up the bill.
AI and the warfare state
In truth, the capitalist system cannot afford to let the AI giants fail completely. Not only are they essential to the profits of the ruling class, but as the world drifts towards greater war and conflict, this technology is taking on a central role in the arms race.
US and Chinese imperialism are now competing for control over advances in machine learning in a new ‘AI arms race’. The question of who possesses the leading edge is an existential question for both powers.
As a result, both are building close links between state power and their respective tech companies. Giants like Palantir are making their integration into the ‘warfare state’ a key part of their growth strategy. On his side, Trump has shown an iron determination to stop any regulations on AI companies which would benefit working-class people.
At root, this is fundamentally about resources and who has access to them. Whoever gains the upper hand in controlling the production of computer chips and the semiconductors needed to make them will have a decisive advantage.
This is one reason why Taiwan has become a central theatre in the inter-imperialist conflict. Taiwan’s Semiconductor Manufacturing Company still produces the vast majority of the world’s most advanced chips. For this reason, it has been a key priority of US imperialism to bring chip manufacturing directly onto US soil.
Both powers are also seeking out new and ever more deadly ways to use AI on the battlefield. On both sides of the war in Ukraine, US and Chinese-manufactured AI-assisted drones have been used to penetrate deep into enemy territory. Microsoft, Google and Amazon have given extensive AI assistance to the Israeli state in its genocide of the Palestinian people.
These developments also raise the horrifying prospect of entirely automated drone systems which can identify targets, authorise strikes and execute killings with little to no human input. While this doesn’t yet appear to be in use, even the possibility of it is a stark reminder of the complete barbarism of this capitalist system.
Should we oppose AI in principle?
AI is not inherently ‘bad’ in and of itself. The question is who uses it, how the technology is developed and trained, and for what purpose.
There are many examples of ways in which AI could be used to drastically improve life. It has been proven that problems in cancer research and drug development that might take humans years to solve have been tackled by non-generative forms of AI in a matter of minutes.
AI also has enormous productive potential that could be used to improve human life, shorten the working week without loss of pay and much more. This technology could also be indispensable in a planned socialist economy – for instance, helping to quickly and effectively calculate what is needed and how much must be produced to meet those needs.
But under capitalism, the impact is fundamentally the opposite. In many ways this is not entirely unique to AI. From the dawn of capitalism, all new machinery – from the spinning wheel to automation – has had a dual potential. What has the potential to improve life also has the potential to worsen it.
This was explained by Karl Marx as long ago as the 1860s. As Marx prophetically argued:
“Machinery, considered alone, shortens the hours of labour. But when [it is] in the service of capital, [it] lengthens them… In itself, it is a victory of man over the forces of nature. But in the hands of capital, [it] makes man the slave of those forces. In itself it increases the wealth of the workers, but in the hands of capital, [it] makes them paupers.”
These words could have been uttered today, and they would speak just as much to the situation we face today.
That’s why it is essential for the working-class movement and communities to expand and step up existing campaigns against big business’s use of AI. Anti-data centre campaigns need solidarity, support, mass mobilisations, and to be linked up and coordinated. Unions in affected industries need to build mass campaigns against deskilling and job losses, fighting for real union protections.
But above all, if the working class doesn’t have full democratic control and ownership of AI, we will not be able to control how it is used.
Fighting for socialism means fighting for public ownership and democratic working-class control of the AI companies, along with much more. We struggle for a society where, rather than using technology to kill and exploit, we use it in the interests of humanity and the planet.
