Responding to Kevin Crane’s recent article – Frequently Asked Questions about AI – Kevin Courtney takes issue on several points in our attempt to develop a more robust analysis of the impact of new technologies

While I share the concerns about AI hype, potential stock market bubbles and the immense difficulties of regulating powerful technology in a capitalist economy, the article mischaracterises what AI systems are actually doing and therefore points us in the wrong strategic direction.

AI is not a ‘stochastic parrot’

Crane, drawing on Emily Bender, argues that AI lacks originality and is merely a ‘stochastic parrot’ – reproducing patterns from human data without understanding or creativity. Human intelligence, by contrast, is portrayed as uniquely creative.

This view echoes a long history of claims for human specialness that have repeatedly been challenged by evidence: the Earth at the centre of the universe, humans as a special divine creation separate from other animals, the supposed uniqueness of tool use or language. Each time, deeper understanding has shown continuity rather than absolute rupture.

The human brain remains the most complex known object in the universe – with roughly 86 billion neurons and hundreds of trillions of synaptic connections. But modern AI systems are closing the gap in scale. GPT-4 is estimated to have over a trillion parameters, with later models growing rapidly. These are not organic, but their architecture is explicitly inspired by neural processes in brains.

Crucially, the ‘knowledge’ in these models is no longer retrieved from the web on demand . It is encoded in the precise configuration of billions of parameters after exhaustive training on vast datasets. This process creates an internal model of the world – not perfect, but remarkably effective. When an LLM (Large Language Model) answers a question, it generates responses based on this learned model, much as humans draw on internalised knowledge and patterns.

Creativity and Originality

The claim that AI merely plagiarises ignores how creativity actually works. Human geniuses like Newton, Marx, or Einstein -standing on the shoulders of giants – did not invent in a vacuum. They synthesised, extended, and transformed prior knowledge. AI does the same, but at scales and speeds previously unimaginable.

Early LLMs did hallucinate and performed poorly on advanced reasoning. Progress has been dramatic. Models have moved from failing high-school maths to acing graduate-level problems. More strikingly, in May 2026 an OpenAI model autonomously solved a long-standing open problem in discrete geometry – the planar unit distance problem posed by Paul Erdős in 1946. Mathematicians had long believed optimal configurations resembled square grids; the AI discovered an entirely new family of constructions that performed better.

This is genuine creativity in any meaningful sense: generating novel, verifiable insights beyond the training data.

Jobs and economic impact

AI is already disrupting employment. HSBC has announced or is planning cuts affecting around 10% of its workforce (approximately 20,000 roles) as part of an AI-driven overhaul, alongside similar moves at other major firms.

Roles that consist primarily of information processing on computers are particularly vulnerable. A smaller number of humans will likely supervise and direct AI systems. Autonomous driving (pursued aggressively by Tesla, Waymo, and others) threatens millions of driving jobs. The pattern is clear: where tasks can be automated profitably, capital will automate them – not to liberate humanity with leisure, but to cut costs and boost profits.

Bubbles and Power Concentration

There may well be a speculative bubble in AI-related stocks, with many companies overpromising. But this does not negate the underlying technological progress. Failures and consolidation will likely result in even stronger monopolies among the survivors.

The Real Stakes

I am not sanguine about AI. Deployed under capitalist logic – private ownership, profit maximisation, and minimal democratic oversight – it risks massive unemployment, surveillance, misinformation, and new forms of control. The same tools that could accelerate drug discovery (as with AlphaFold-style systems), scientific breakthroughs, and cultural production could instead immiserate workers and entrench power.

The fundamental argument is therefore not ‘AI is just a parrot, don’t worry’ or uncritical boosterism. It is that this technology strengthens the case for social ownership and democratic planning of the economy. We need systems that direct AI toward meeting human needs – curing diseases, expanding knowledge, reducing drudgery, and expanding genuine leisure and human development – rather than private accumulation.

That requires an accurate understanding of AI’s capabilities and trajectory. Underestimating it leads to poor strategy. The left should engage seriously with the technology’s strengths and dangers, not retreat into comforting but outdated dismissals.

The original Counterfire piece is a useful starting point for debate, but we need sharper analysis if we are to shape AI’s development in the interests of the working class and humanity as a whole.

Kevin Courtney is the Chair of the Cuba Solidarity Campaign, Interim Chair of the Together Alliance and was previously the Joint General Secretary of the NEU. He writes in a personal capacity