Groucho Marxism

In the last few years there has been a rapid adoption of artificial intelligence across all sectors of society. This has been driven largely by the development if so-called ‘generative AI’ technology. Generative AI is a subfield of artificial intelligence that uses generative models to produce text, images, videos, audio, software code, and other forms of data. Unlike traditional AI that analyzes data to support decisions, generative AI produces novel outputs in response to user prompts, using models trained on vast datasets. Firms have been quick to jump on the generative AI bandwagon, mainly through fear of being left behind by their competitors. As exciting as this new technology seems, however, there are many potential downsides.

The most obvious downside of the mass adoption of AI is often referred to euphemistically as ‘labour market disruption’, but is more accurately referred to as ‘mass unemployment’. This is something that inevitably occurs under capitalism with the widespread introduction of a labour-saving technology, and there are many historical case studies going back to the Luddites of the early 19th century and beyond. Usually, though, it is manual workers who are most at risk. What’s both interesting and novel about the current generative AI boom is that those most at risk of unemployment are high-skilled, white-collar professionals, such as writers, analysts, and legal assistants. I don’t think many people saw that one coming.

At this point it is worth reminding ourselves of why unemployment is a problem in the first place. Unemployment being a bad thing is usually taken as axiomatic – that is, something that does not require any explanation. But if you think about it, it’s bizarre that our first reaction whenever a new labour-saving technology is introduced is to worry that it might mean us humans have less work to do! Shouldn’t that be a good thing?! The reason it isn’t a good thing is that we live under a system which requires 99% of people to sell their labour-power in order to survive. Under capitalism, if you are unable to sell your labour-power, you are in deep trouble. Labour-saving technology is seen as bad as it inevitably results in less opportunities for us to do this.

This isn’t the fault of the technology though; it is the fault of the system we live under. What makes this even more perverse is that under capitalism, a lot of people have jobs where the main objective is to devise ways to reduce human labour! We should not be surprised by this though; it is just one of the many contradictions of capitalism, and these contradictions have existed for as long as capitalism has. However, the rise of generative AI has brought this contradiction into sharp focus. In the past, we’ve always been able to invent new jobs for people to do – even if a lot of them are ‘bullshit jobs’, in the memorable phrase of the late, great American anthropologist David Graeber. But the proliferation of AI is going make it increasingly difficult for us to invent jobs that can’t be done by a machine.

There are other downsides to generative AI too. As these models are trained on vast datasets, they often perpetuate existing societal biases and stereotypes, leading to discriminatory outcomes in areas like recruitment. The ability to easily generate convincing, but false, information (hallucinations) and malicious content like deep fakes poses a threat to information reliability and security. We are not quite at the point where AI-generated content is indistinguishable from the genuine article, but we aren’t far off. Another major downside is that training and running large AI models require substantial energy and water consumption, which will inevitably exacerbate the climate crisis. Then there is the problem of over-reliance on AI leading to the erosion of essential human skills such as critical thinking.

We would do well to remember that for all its supposed intelligence, generative AI is just compiling information originally created by human beings. This information is often obtained from specialist online forums. In future, generative AI might make such forums obsolete, so the information will no longer be created – and what happens then? Presumably, AI will start creating content based on information it has previously collated, effectively turning itself into a self-replicating feedback loop. It might be tempting to speculate that AI might then start to evolve and even become sentient, as self-replication is the key mechanism behind evolution by natural selection. Indeed, many people are worried about exactly such a scenario occurring.

Personally, though, I think these fears are overblown. Evolution by natural selection requires more than just replication; it also requires a constant stream of new information. This is essential for the ‘selection’ part of the process to work effectively. As we have seen, new information would not be present under the scenario just described. A more likely outcome under this scenario is that generative AI ends up becoming irrelevant and simply stops being used. Given all the downsides of this technology, that might not be such a bad thing.