Song-Chun Zhu, in the field of Artificial Intelligence, is taking a strikingly different stance from the prevailing paradigm in the US, reports Chang Che in The Guardian Weekly (26 September 2025). He argues that a sign of true intelligence is the ability to reason towards a goal using minimal inputs. He contrasts this with the ‘big data, small task’ approach employed by large language models.
His critique makes sense to me. Finding patterns and connections in existing data works with what we already know or think. But if the future emerges in ways that are ‘shaped but not determined’ by the past – where entirely new things may emerge – then analysing the past, with whatever depth, aggrandises ‘what is’ and ‘what has been’. It is normative and also preferences the idea that the future is a smooth extension of the present. And, by preferencing conclusions on that basis, it limits how we imagine the possibilities of what could be.
I have no idea how AI could do ‘small data big task’, and I am not writing (in case you were in any doubt) in support of handing over even more power to machines. But I think Zhu makes a good point. Creative people have an ability to make sense of complex situations through using intuition and working with/valuing subjective and qualitative data. And out of that messy interconnected data we can often see ‘simplicity on the other side of complexity’, a topic I explore in The Dao of Complexity. By not making a priori decisions as to what is relevant and where are the boundaries, we are more likely to notice key dynamically stable features or spot just-showing new shoots or nascent ‘cracks’ in the fabric. We may sense into meanings, coherences, emergences which don’t just show themselves through analysing what is and what has been. Those creative intuitive leaps are what characterises scientists such as Einstein or Lovelock or Darwin and exemplify the approach of many entrepreneurs and inventors.
Post-script
Incidentally, this idea that AI can solve any problem was exemplified in an article by Krotoski (Guardian Weekly 3rd October 2025) about the “techno-fundamentalist dream of AI-enabled immortality” where “ageing is a technical problem that can and will be fixed.”
He goes on:
Too much faith in data and engineering overlooks the value of the unknown and unknowable… This impulse, this motivation this moral mandate to want to improve yourself means we have to become like a machine.. [Yet] we are weird, we are messy.
Image credit: Rich Armington