W. Brian Arthur speaking at The Enlightenment Conference marking the 250th anniversary of Adam Smith’s (1776) “Wealth of Nations” inside Smith’s Panmure House in Edinburgh.
One of “the old guys” speaking on subject matter that already seemed old when I was doing my Master’s in the late 1980’s, the early days of the “digital revolution” (Even back then, we were studying robot-&-video-assisted, minimally-invasive human surgery, already happening, courtesy of HP-PARC if I recall correctly).
Anyway the subject then (as ever) was technology driving economic activity and change, specifically “computer” technologies then and now. Arthur calls it “digital”, I call it “Information Processing” (it’s not all digital), partly a nod to the fact that Smith’s more important work was his earlier (1759) “Theory of Moral Sentiments” how we think and decide to act for the best, individually within collective “markets”. My “Complex Adaptive Systems Thinking” cybernetics topic to this day.
Despite that old-hat impression, we are in fact getting Arthur’s wisdom. Throughout my course and dissertation, it was a given that technological change was (and is) the main driver of economic change - the way people spend their lives adding value to the world - much more so than changes in financial and human capital. In typical management theory style - control by pulling levers - not much thought was actually given to the true relationship between technology and the economy - you invented it and then you adopted & exploited it. Arthur’s point is that the economy “encounters” new technology and its possibilities bubbling up from wherever its invented - like volcanic magma - not that the invention is caused or incentivised by existing business. The encounters are at all levels in the management, operational, market and supply chains, quite independent of where the invention occurs. His volcanic metaphor continues to suggest that rather than simply adoption, its more like combining business with the technology and reifying into “new lands” which become the new economy.
Back in my day, the main theories on how technology created economic change and revolution (in that broadest sense) were Schumpeter’s “Cycles” or Kondratiev “Waves” and of course Thomas Kuhn’s “Paradigm Shifts”. Focussing on the big-picture, where the causal relationship between technology and economics was a presumed given. It was Carlotta Perez’ multiple contributions in Dosi, Freeman et al that coined “Techno-Economic Paradigms” combining those cycles into new lands. Arthur strongly recommends Perez work generally, and I see now that Arthur was another single contributor to Dosi, Freeman et al. Small world.
Hat-tip to Dave Snowden on LinkedIn for sharing the Arthur lecture and his own thoughts. In 2026 the main focus of the “digital” paradigm is of course AI.
“ Listening to a talk by Brian Arthur (fond memories of conversations over meals in Singapore and, most recently, the King’s Head in Oxford), this is just brilliant (paraphrasing): “As kids, we used to go on ‘mystery tours’; the driver knew where the bus was going, nobody else did. Generative AI has us on a mystery tour too, except this time the driver doesn’t know where he’s going either.”
Interesting dialogue arising between people suggesting that actually “there is no driver” and people like me suggesting the real drivers are the tech-bros, who are sadly misguided on how the world works and what it needs and scarily overpowered in their dysmemic reach and influence (even though not overall control) compared to the free-democratic nation states intended to provide our governance.
[Post Note: And Dave being as prolific as he is, despite being just out of a cancer surgical procedure, has already written-up his thoughts.]
Compare also this recent Grauniad opinion piece by Timothy Garton Ash “Would even an AI disaster on the scale of Hiroshima be enough to make humankind protect itself? I fear not” (hat tip to Dennis Findlayson). Largely reinforces my own positions:
A. Whilst LLM-AIs are tools, the problem is humans with too much power and too much confidence in how they themselves understand what the world needs (ie the Tech Bros). That’s why my systems focus is our individual and collective governance (ie the original aim of cybernetics - better democracy).
B. The “existential disaster” scale risks are when AI itself actually becomes “intelligent” and as this piece says (and I say at least once a week for the last 25 years) that happens when it can evolve itself, ie AI becomes A-Life with it’s own life goals. We should prevent that at all costs.
C. Between the two cases as I say the problem is the vicious cycle of people A with more power and wealth than the nation states we use to govern ourselves to prevent case B and thereby enable it. Catastrophe’s are “predictable” at many levels and scales, from biases within AI platforms (Eleanor Drage’s feminist AI) , to the ”singularity” end-game for humanity. The response is better governance and requisite diversity now.
There’s more useful metaphors in Arthur’s piece of wisdom, worth a watch, beyond my Psybernetics agenda here.
=====
Post Notes:
And, today Steven Pinker too:
“If data center construction were to stop, it wouldn’t just reduce growth by 1 percentage point. It would also turn companies’ optimism into pessimism, as they realized that the AI boom was going to be throttled by NIMBYs and populists. ...
Changing optimism into pessimism is a classic ‘animal spirits shock’ or ‘expectations shock’ — something that reduces aggregate demand. If hyperscalers and AI-positive companies and equipment manufacturers stop spending big, there will be fewer dollars going to construction workers, engineers, and many other workers. ...”
Animal spirits there - Adam Smith and Robert Pirsig - bio-social relative to formal intellectual considerations.
And, a more upbeat AI thought from the Ontolog-Forum about the “autistic” cage of Tractatus style logical propositional forms of knowledge, is that in the right hands (see, not the Tech Bros), implementations of language model AI’s don’t need to be bound by such syntactical constraints. We could actually have AI’s that “understand” any kinds of human expression. I need to get to know Dennis de Champeaux.
=====