 When the Laboratory Feeds the Machine: The Commoditization Stack and the Cross-Border Appropriation of Open Science: Autonomous Driving as the Mirror CaseArthur de Miranda Neto ()
Working Papers from HAL Abstract: This is not a recommendation, but an invitation to reflect on the arguments presented in this paper. Open science rests on a compounding premise: that knowledge released without restriction returns, at least partially and in various forms, to those who produced it. For most of its history this held, in part because the maturation of knowledge was slow enough that returns accrued to producers before value could be captured elsewhere; it was a fragmented system that worked because time was on society's side. This paper argues that, under generative-AI commoditization, the premise fails asymmetrically. Open science has been among the most productive arrangements in the history of research, and its productivity depended on collaboration, because knowledge was distributed across specialists; that condition is now changing as capability concentrates in a few compute-rich frontiers and is commercialized; the argument is therefore not against openness but an invitation to reflection that the new conditions make timely, offered so that what has worked can be preserved rather than assumed. The SAE J3016 taxonomy of driving automation serves as a scaffold, crossed with a seven-layer commoditization stack, so that autonomous driving can be treated as the mirror case of AI-era knowledge production: a domain whose decomposition into perception, localization, mapping, planning, and control renders commoditization measurable subsystem by subsystem; the case is a lens rather than a limit, for the same mechanism recurs across many other knowledge-intensive activities. Two regularities are established. First, artificial intelligence, and increasingly agentic artificial intelligence, penetrates the stack top-down, with penetration depth inversely proportional to safety-criticality and certification burden; the defensible residual therefore concentrates in judgment and operational-design-domain (ODD) setting (Layer 5) and in certifiable accountability (Layer 6). Second, the agentic design now proposed for driving coincides with the one proposed for scientific discovery, which makes driving and science two instances of one mechanism. The cross-border knowledge regime, captured by a coefficient K7, is then reframed. Open scientific output is ingested unilaterally into frontier models held by two poles, the United States and China, stripped of attribution, and re-exported as priced capability, so that value need not return to the producing laboratory or nation despite heavy investment. Decomposing K7 into a collaboration component and an ingestion component shows that decoupling in collaboration need not reduce exposure to appropriation. The same fragmented frontier also gates the defensible residual. The certification and verification capability that Layer 6 requires is bloc-held, leaving a non-frontier actor unable both to certify its agentic build and to defend it. A real but anonymized case, two internationally leading European public laboratories, instantiates the argument, and the pairing of an engineering laboratory with a social-science laboratory is treated as methodologically necessary. A concluding contribution, offered as a matter for reflection, asks whether the knowledge most exposed to appropriation might be converted, within universities, into the formation of people: an asset that resists ingestion and that may help graduates withstand the difficulties of the first job. Keywords: operational design domain; higher education; sovereignty; cognitive capitalism; data colonialism; certifiability; cross-border knowledge regime (K7); open science; agentic AI; autonomous driving; SAE J3016; commoditization stack; generative AI (search for similar items in EconPapers) Downloads: (external link) Related works: Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text Persistent link: https://EconPapers.repec.org/RePEc:hal:wpaper:hal-05702905 Access Statistics for this paper More papers in Working Papers from HAL |
Is your work missing from RePEc? Here is how to contribute.
Questions or problems? Check the EconPapers FAQ or send mail to .
EconPapers is hosted by the
School of Business at Örebro University.