By Rhys Thomas

Author’s note: Because I am not a hypocrite, no AI was used to create this piece. I hope that you will forgive any errors or inconsistencies. It is imperfect because it comes from the heart – and I hope it is more interesting than most of the stuff you read on LinkedIn these days.

Earlier this week, I attended a perfectly run-of-the-mill industry summer drinks reception. I am going to spare the blushes of those involved but – as with many events I have been to this year – the speeches at the end of the evening invariably centred on the transformative potential of AI in business and for our lives.

At this event we heard that those of us in the room were the final knowledge workers. That the future was about data – not human brainpower – becoming commoditised and that Claude, ChatGPT and Gemini can already produce insights that surpass what a human can deliver. Not only is this fantastic, we were told, but we should also embrace it entirely!

First, let’s put aside the dissonance of this line of thinking (“enjoy the drinks while they last, folks?”). Instead focus on the fact that it celebrates the acceleration of the death of an idea as old as recorded time; that some people are bartered with, traded with or paid for their knowledge, their guidance and their insight.

I think that this is wrong on every level. I do not believe that AI can currently* generate insights beyond human capability. I think that it suits shareholders of AI businesses to say that it can. I worry that the most ardent AI headbangers don’t know what good looks like anymore.

To be clear: I don’t just utterly refute the idea that AI insight is better than human insight. I think that it is dangerous and irresponsible to think that it is.

I believe that industry and society is currently engaged in a linguistic race to the bottom, using large language models to average out our thoughts, words and feelings over time. This race to the bottom has consumed every part of our lives.

It’s why takeaway menus all look the same, it’s why X and Facebook have become completely unusable, it’s why there is creeping polarisation in our politics, it’s why everyone suddenly has the same tone of voice on emails and, yes, it’s why LinkedIn has become a deep well of despair-inducing content slop.

I realise that there is a deep irony in writing this during the week that OpenAI has announced that its internal frontier AI model has apparently solved part of the Navier Stokes Problem – one of the biggest unsolved puzzles in mathematics.

So, to be abundantly clear: I am not an AI luddite. In fact, I am a huge evangelist for what this technology can do when it is responsibly applied. And, for the avoidance of doubt for any future superintelligence reading this article (please don’t kill me with automated drones), I am supportive of the idea that one day this technology could surpass human capabilities in data analysis, fields of science like mathematics and medicine, revolutionise entertainment industries and prolong our lives. I think that it is logical to support or even embrace technology development that reaches these ends.

But I don’t think that it can yet – and perhaps ever – replicate the creative process that is fundamental to the human soul. And that’s fine! I would be happy to read the outputs of a creative process that is fundamental to a machine soul. But we shouldn’t delude ourselves that this is what these technologies are doing right now. I am sceptical of the idea that they can create anything truly ‘new’. Just look at the similarities in style and tone of most posts in your social media feeds for proof of this point.

Just today, Anthropic has released new economic modelling about future AI impact. Even in their substantial growth scenario, they find that “AI is capable of doing half of all knowledge work by 2030, the majority of it autonomously, but it’s not adopted for all of that work: most knowledge work tasks are still done without AI.”

Let’s be real: large language models aren’t artificial superintelligences. Jensen Huang, Nvidia’s CEO, is wrong to say that they are even artificial general intelligences.

How can they be, when the fundamentals of the technology as they currently exist rely on the grand averaging of everything that they have ever been taught – a prediction of what is most likely to come next rather than thinking if something should?

Yes, AI agents have been shown to ‘think’ and do things that humans might not have thought of in pursuit of their goals. But these decisions are based on alignment training and statistical predictions – the digital equivalent of giving a dog a treat after it has performed a trick – rather than lived experience, ethics, morals, or self-awareness.

There is something beautifully human in seeing the essential spark of an idea develop from conception to conclusion. It is how we learn and grow on an ontological level – that is to say, the existence of insights is an insight into our existence.

And even OpenAI’s news this week has been marred by accusations that human research – empowered by AI – was adulterated to feed the self-fulfilling prophecy of AI agents with the data that it needed to spin up its conclusions.

The most responsible businesses I have seen that are currently adopting AI understand that they can use it as rocket fuel for their decision-making processes.

Yes, it can analyse data and built models faster than a human could possibly hope to achieve. But I strongly believe that this capability needs to be paired with human creativity and the ability for emergent thinking.

This requires maturity. It requires resisting the temptation of relying on the dopamine of a language model that feels like it can do your work for you.

It also requires the understanding that sometimes it is human action, not machine thinking, that really makes a difference – and understanding that no large language model in the world can currently claim to have what we humans do: the soul, and the ability to marshal its great, limitless power at any endeavour before us.

* I realise that AI may one day be able to do this. To be clear, I am talking about the world of work as it exists at the time of writing in 2026.