Artificial intelligence, from tool to institution

The latest Nobel Prize in economics awarded to Daron Acemoglu and colleagues draws our attention to (and urges us to) a dimension that is currently overlooked and underestimated (but highly relevant in perspective) in the current debate around Ai: the institutional dimension. That is to say, the fact that the new artificial intelligence engineers are not just mere efficient tools but rather and potentially surprising resources and institutional forces. When read in this light, AI technologies speculatively take on a far more radical and refoundational scope, we might say, than a banal instrumental approach would allow. Take, for instance, large-scale language models (ChatGpt et similia). We can interpret them 'instrumentally' as sophisticated (a) text predictor, word calculating machines. Eminently rhetorical techniques designed to predict the word (and speech) most likely to follow a textual sequence given as input. We would not be wrong to read them this way, but there is more.

These tools, once introduced in companies and organisations, will transform modes of production and divisions of labour, cognitive dynamics and competitive strategies. In this second reading, from mere text predictors they 'structurally' become (b) knowledge engines, knowledge engines capable of subverting the classical modes of information creation, circulation and storage. Thus, from a mere instrumental 'solution' they become a new industrial 'system'. And yet, that is not all. There is a third, perhaps surprising, way of reading the llarge language model. And that is to consider them, more 'institutionally', as (c)market makers. That is, socio-technical devices capable of giving life to new, more-than-human institutionalities by disarticulating and refounding other more classical institutions such as markets, bureaucracies, states, due to the arrival of renewed institutional logics of coordination, communication, transaction, decision-making, coercion and so on. Of course it may be alienating for us to consider large-scale language models as (neo)institutionalities, but we should not be so surprised. Markets too - economists say - have the same institutional (institutionalising) capacity to process collective information with supply and demand (and they too, like the large language models, combining intelligence and stupidity) to determine collective actions and decisions. Let us further think about what will happen with the arrival of autonomous artificial agents and the definitive transformation of machines from mere instruments of production to sophisticated economic actors with machines consuming services, but also creating enterprises (someone has already renamed them robopreneurs).

Talking and transacting machines together (agent-to-agent system) therefore, not without vulnerabilities and risks. Once endowed with operational-cognitive (ai agents) and ultimately operational-financial (crypto wallets) autonomy, machines with 'artificial minds' and 'digital wallets' will radically transform the institutional nature of markets. And perhaps 'market' at that point will no longer even be an adequate term to describe the new economy. In the speculative light of this, an epistemic twist from innovation to institution is then required: from textual predictors to structural engines to institutional devices. It is a political and anthropological, legal and regulatory change of step as well.

From considering Ai an existential risk to reading it as a potential (risky even) institutional resource. Thus, the more the power of computation scales planetarily and enters the economy (from the code economy to the machine economy to the agentic economy), the more our strategic gaze must shift from 'instrumental' (and industrial) to more radically 'institutional'. From corporations to markets, we do not govern and prosper within this 'supertransition' without a speculative and philosophical shift to the institutional thinking promoted by the new engineering. Which are as much terrestrial as they are orbital. Even cosmic, we would say, given the progressive competitive extension of economics and politics into extraterrestrial space.