Mark Best, Socialist Party National Committee
The latest developments in Artificial Intelligence (AI) – the impact they might have on workers’ jobs, livelihoods and our day-to-day lives – are being discussed in the press. Recent advances don’t mean we are about to develop machines capable of human or superhuman thought, but have brought forward the potential automation of a wider array of jobs and tasks. The race between different companies to produce ever bigger and more powerful models in the hope of capturing the lion’s share of future profits has some capitalist politicians worried about the economic, social and political impacts these new technologies may have.
At a time when capitalism across the world limps from crisis to crisis, what consequences could AI implementation have and how do we fight to make sure the potential gains of new technology don’t come at the expense of workers’ living standards?
The impact of new technology isn’t inherent within the science itself but will be determined by the outcomes of battles between workers and the bosses, dependent on the organisation of both sides. How could society be organised in a way that science and technology is developed for the good of us all instead of what makes a profit for a few at the top?
What can the tech do?
Artificial Intelligence in the broadest terms refers to machines and computer programs able to react to conditions and complete tasks. This includes computer programs that can play chess, recognise images or check your spelling and grammar, and even, in a way, many of the mechanical systems that make up robotic production.
The explosion in the capability and availability of generative AI algorithms – programs which when given an input will output text, images, computer code, music and so on – has thrown AI and its implications for the capitalist system under the spotlight.
Open AI’s ChatGPT, Google’s LaMDA and other generative algorithms can take part in human-like conversation, produce realistic photographs and video, and write convincing articles.
Algorithms able to mimic human writing and art may give the impression that there is a thinking, intelligent being behind them. But there is still a large jump to so-called Artificial General Intelligence (AGI) with the capability to think and reason as humans can, and beyond.
AI technologies are developed by machine learning; algorithms that have the ability to modify themselves, to ‘learn’ by checking if their output matches what is expected and making internal changes to approach an ideal response.
In the case of the large language models that produce text, vast amounts of written material from books, text from the internet, computer code, etc. have been plugged into a model which processes and refines itself. Similarly, image generators have been trained with thousands of images. Advances in computing, allowing for more processing of testing data, have enabled these powerful algorithms to be developed.
These algorithms have been trained using the results of hours of human creativity, sometimes seemingly without the permission of the creators themselves. Some image-generating algorithms produce new photos with watermarks of popular stock image sites already baked in, hinting where the training data may have originated from!
Hollywood’s ongoing Writers Guild of America strike for better pay and rights includes demands related to generative AI. It calls for a ban on writers being made to work on or edit scripts created by generative AI, and to not allow language models to be trained on material produced by scriptwriters.
It’s not just in the creative industries, the automation of a number of jobs in a wide range of fields is posed. Goldman Sachs, the investment bank, predicts that two thirds of jobs in Europe and the US are exposed to some degree of automation and that a quarter of work could be replaced. It estimates this would mean 300 million full-time jobs automated by algorithms, with an unknown number of new jobs created as the technology is implemented.
Although the exact impact new avenues of automation will have on jobs is unknown, individual capitalists are compelled to shrink their wages bill to look after their profits and keep up with their competitors.
Responding to lower than expected profits, BT has announced it plans to cut 55,000 jobs from its global workforce by 2030. While many of these come from the ending of the rollout of fibre broadband, the company promised 10,000 of these jobs will be replaced by generative AI. Is this technically feasible or does it represent an attempt to reassure investors that although BT is not profitable now, AI will make it so in the future?
BT workers in the Communication Workers Union recently took strike action, extracting an improved pay deal from the bosses. And attempts to make tens of thousands of redundancies in the future will be met by resistance from a unionised workforce.
Mass job losses won’t develop without resistance from the working class. Any attempts to make workers pay for advances in new technology must be resisted by collective action. Trade unions should demand that gains in productivity aren’t used to sack workers or drive down wages. The demand should be put on bosses who argue such measures are necessary to open the books to trade union inspection.
Companies which proceed with attacks should be nationalised to save jobs, with compensation only on the basis of proven need. Publicly owned and run democratically by the working class, measures can be agreed to share out remaining work without loss of pay. High-quality education and training on full pay, and with a guaranteed job at the end for those who want it, must be fought for too.
Entry level, less-skilled jobs will be at risk of automation. Many already are with the introduction of things such as self-service checkouts. But traditionally better-paid professionals, which the capitalist class has been able to draw some social support from historically, losing their jobs to automation can further undermine support for the capitalist system. Already, attacks on pay and conditions have forced lawyers and doctors to take strike action, pushing them closer to working-class methods of struggle.
For some sections of the capitalist class, including some of those involved in the development and monetisation of AI, the risks of continued unregulated, unplanned advances in AI come in many forms. Individual capitalists worry their capital and ability to make a profit may be adversely impacted by new developments, or they may have already missed the boat and the existing AI industry leaders have already won out.
These developments are another source of instability in an increasingly chaotic global situation.
Already malicious computer programs have caused real damage to physical infrastructure. The Stuxnet virus which spread around the world is believed to have been designed to disrupt Iran’s nuclear programme, reportedly damaging up to 1,000 centrifuges. Computer programs generating viruses, either by design or accidentally, could have real economic consequences.
Automated stock-trading programs already exacerbate instability in the financial markets, in some instances causing stock values to rapidly collapse and recover within minutes. Far from creating these crashes, new technology has merely sped up transactions, feeding into stock markets used by bankers to gamble trading increasingly complex things divorced from economic reality.
Calls for regulation from the developers of these AI research companies have been presented as attempts to minimise the ‘damaging impact’ the new technology may have. It is actually primarily an attempt to lock in the handful of companies currently ahead in research as the leaders of the field, with the profits that will come with it. States, acting in the interest of their nations’ capitalists, may regulate the development of AI to protect their position and profits on the world scale.
AI models will continue to be developed and released in an unplanned way in attempts to capture parts of the market, as has already been done. This is the way the capitalist system works. The drive for profit pushes individual capitalists down this road, even if it has the potential to destabilise or damage the capitalist system as a whole. As Geoffrey Hinton, called a ‘godfather of AI’, who recently resigned from Google and warned about the impact of AI, said: “The problem is, in a capitalist system, if your competitor then does do that [develop and release larger, more powerful models], there’s nothing you can do but do the same.” The direction AI is heading, as has happened with new technologies in the past, is of a few companies monopolising knowledge under lock and key, stifling the free and open development of science and technique.
Capitalist automation
Over the course of its development, capitalism has enabled the replacement of manual jobs in production with machines. Driven by higher profits, capitalists replaced workers with machines, implementing advances in technology to compete with rival producers. Through the socialisation of production, advances in science and technology were able to be used to take society forward and improve overall living standards, while increasing the gap between those at the top and bottom of society.
This trend, of increased automation in production, cannot be taken to its conclusion by capitalism. Capitalists only invest in implementing new technology when they can be sure of realising a profit. This means that new technology will be implemented if the capitalists think society is stable enough to justify investing large amounts of money now to realise a profit over a number of decades, and if there aren’t other more profitable opportunities for them to use their capital in the short term. The production of semiconductors, a vital component in the chips in all modern electronics, requires $15-20 billion of investment for a new factory. Capitalists have to be persuaded to invest in new infrastructure with government subsidies, even though profits for these are likely to be guaranteed over their years-long lifespan.
Already automatic factories, in which raw materials enter and finished products leave, exist. That they are not implemented on a wide scale indicates that in many industries paying and attempting to drive down workers’ wages as far as possible is more profitable than full automation.
The struggle over the impact of new technology is class struggle, with workers fighting for jobs and wages on the one side and the bosses who own the factories, the patents and the infrastructure on the other. Under capitalism, technological advances will be used in ways to maximise profit.
A democratically planned socialist society, based on public ownership of big business and the banks, would be able to use technology in a different way. Instead of hunting for profit, production could be planned with other priorities in mind, such as reducing the necessary time for people to have to work, alongside protecting the environment and the wellbeing of workers.
In a number of fields, the potential for technology to be used for the good of us all is there. The advances in electronics and computing which have fuelled the gig economy contain within them the potential to be used to improve the lives of workers across the world.
Automating jobs with generative algorithms could mean some of the drudgery of life is abolished. Workers in call centres, admin staff processing forms and workers doing jobs only necessary for the maintenance of capitalism, could have a real choice about what they want to do. Whether that is to use their current skills or to do something completely different.
A socialist society, in which the resources and knowledge of society are owned and directed democratically by workers themselves, would mean technology is freed from the narrow interests of the capitalists. Technology which under capitalism condemns workers to lower wages and unemployment could be utilised for the good of us all.