AI can search through enormous spaces of possibilities, but this paper argues that searching is not the same as discovering. The most entrepreneurial opportunity may be the one nobody was looking for. ‘Artificial Intelligence, Entrepreneurial Discovery, and Embodied Tacit Knowledge’ is presented as one of the latest steps in the long ‘socialist calculation debate,’ which argues back and forth over the possibility of central economic planning.

The most recent developments of that debate revolve around the impossibility for a planned economy, whether AI-powered or otherwise, to recreate entrepreneurship. Kurun identifies five epistemic functions of markets and concludes that the last one, entrepreneurial discovery, is impossible to replicate at the present state of the art of AI.

The quest to satisfy everybody’s economic preferences is a moving target because human wants shift unpredictably. In a market economy, new demands are discovered by alert entrepreneurs, whose function is to think creatively and bring about a new state of the economy by using available assets in a novel way to introduce innovation.

Alertness is the capacity to find misallocations with respect to potential new consumer demands that are currently unmet, and that alertness cannot be reproduced by AI because it is a situated performance requiring affective and sensory coupling and therefore embodiment.

Kurun’s argument echoes Dreyfus’s argument about AI’s lack of material participation in the lived experience that shapes human decisions at a deeper level. Embodiment enables entrepreneurs to be sensitive to the economic affordances emerging in market contexts, which are unique to human beings when potential consumers are human beings.

Alertness is a practical virtue resting on know-how that cannot be expressed in propositional knowledge and therefore cannot be exercised by disembodied cognition. Disembodied cognition may at best optimize existing allocations of resources, not deliver undiscovered opportunities that are not implicit in any body of training data.

Discovery and search should be distinguished. Search is a Robbinsian activity: it presupposes a known probability distribution over possible finds, a known cost of search, and optimization of the trade-off between search costs and expected returns.

Discovery, by contrast, occurs when something is noticed that was not being looked for: the unplanned perception of an opportunity whose existence was not previously suspected. DABUS is designed to explore recombinations of ideas that are preselected as relevant or promising by a human mentor.

In that sense, the human mentor performs the entrepreneurial function of identifying viable opportunities. DABUS offloads to the operator the acts of fixing the probability distribution of the situation and considering acceptable costs of search and trade-offs.

LLMs can lower search costs, compress the time needed to survey existing market conditions, and assist entrepreneurs in brainstorming identified opportunities. But lowering the cost of search is not performing discovery: it accelerates exploitation of opportunities already recognized while leaving the prior act of recognition to the embodied, institutionally and historically embedded human agent.

Kurun says the conclusion is not unheard of, but that the diagnosis and weight of the problem receive originality in his paper. The core feature that AI misses to become entrepreneurial is the perceptual basis on which economic opportunities are recognized. Martinelli instead argues for the same conclusion because AI systems lack teleological autonomy: the capability for setting new goals that are not implicit in the current state of the environment.

That ontological strategy is counterbalanced by the rejoinder that a sufficiently autonomous system, an AGI with genuinely original goals, would close the gap, leaving the challenge of arguing that AGI is impossible in principle. The paper is presented as an excellent steppingstone in the socialist calculation debate.

It offers an important and interesting angle on the link between entrepreneurship and the market order. The argument highlights the importance of entrepreneurship for fostering innovation and the dynamic evolution of the economy. The embodiment approach is judged persuasive, while other ways to reach the same conclusion remain valid and reinforce the conviction.

The central takeaway is that AI may optimize known possibilities and reduce search costs, while entrepreneurial discovery depends on embodied, situated alertness. That distinction matters for debates about markets, innovation, and economic planning.