AI models can ship a company's hard-won know-how elsewhere instantly — something human expertise never does. What actually travels, and what gets left behind?
Does machine tacit knowledge truly travel across organizations where human tacit knowledge stays local?
This explores whether the know-how built into AI models really spreads across companies in a way human expertise doesn't, and what gets lost or stays behind when it does.
This explores whether the know-how built into AI models really spreads across companies in a way human expertise doesn't, and what gets lost or stays behind when it does. The short answer from the corpus: the model's know-how travels, but the meaning doesn't come with it. Machine tacit knowledge can be shipped at scale once it's inside a model, yet it only becomes useful knowledge inside a company after people and the system go back and forth enough to build shared context Can machine tacit knowledge spread across organizations unlike human tacit knowledge?. So what travels is a capability. The organizational knowledge still has to be grown locally.
The less obvious part is where that portable knowledge comes from. It started as human tacit knowledge. During RLHF, human raters' gut judgments about which answer is better get turned into explicit rankings. Those rankings train the model, and the model then shows up inside thousands of other companies through products and APIs Where does ChatGPT's tacit machine knowledge actually originate?. Human tacit knowledge does stay local, but local to the AI lab, and the lab is the one place where it gets packaged for export. It's a fair question whether the model actually 'has' tacit knowledge in a strict philosophical sense. There's some early causal evidence, but it rests mostly on a single fact-editing study that has proved hard to replicate Do language models possess tacit knowledge in Davies' sense?.
The contrast with humans also isn't as clean as the question suggests. An industrial case study wrote experts' rules and design principles into an AI agent's scaffolding (the instructions and tools wrapped around the model). Non-experts using it produced work that experts rated at their own level, and output quality rose 206%. The gain came from writing the expertise down in the scaffolding, not from a bigger model Can codified expertise let non-experts match specialist output?. MetaGPT shows the same thing for teams of agents: human standard procedures, written into structured documents the agents pass to each other, coordinate better than open-ended chat between agents Does structured artifact sharing outperform conversational coordination?. Human know-how can travel too, once someone does the work of writing it down.
What gets lost on the way may matter most. One line of thinking says AI moves knowledge around as a continuous flow, like oral cultures did, but without the speaker or giver who used to vouch for it Is AI returning knowledge to flow-based economies?. A sharper version says AI output works like hearsay: secondhand, changed with every retelling, and impossible to trace back to a source Does AI-generated knowledge have the same structure as hearsay?. Farrell and colleagues call large models lossy summarizers of human culture, closer to markets or bureaucracies than to minds Are large AI models actually intelligent agents or social technologies?. Taken together, these notes say machine knowledge travels because it has been cut loose from the people and situations that gave it meaning. That's why each company receiving it has to rebuild that context for itself.
Sources 8 notes
AI introduces tacit machine knowledge that, unlike human tacit knowledge, can be distributed at scale once embodied in models. However, it becomes organizational knowledge only through human-machine feedback loops that build shared context and meaning.
RLHF training replays Nonaka's tacit-explicit cycle within AI laboratories: human raters' tacit judgments become explicit rankings, which train models whose outputs reveal new capabilities. This lab-internal spiral produces portable tacit machine knowledge that feeds knowledge spirals across countless organizations via software products and APIs.
Transformer LLMs can meet Davies' criteria for tacit knowledge based on architectural features and causal tracing with ROME edits. However, evidence rests on a single fact-editing case, and replication challenges suggest the causal localization may not be as precise as initially claimed.
An industrial case study embedding domain rules and design principles into an LLM agent's scaffolding achieved 206% output-quality improvement and expert-level ratings from non-experts, bypassing the need for specialist oversight. The capability gain came from externalizing tacit expertise into structured harness components, not from model scale.
MetaGPT demonstrates that agents producing standardized engineering documents achieve superior coordination compared to conversational exchange. Active information pulling from shared environments eliminates noise and mirrors efficient human workplace infrastructure.
Show all 8 sources
Print culture fixed knowledge as accumulated stock; AI returns knowledge to generative flow. However, unlike oral and gift economies, AI flows lack the embodied transmission—the speaker, the giver—that historically anchored knowledge circulation.
AI output shares all defining features of hearsay: testimony at remove, modification in retelling, unattributable origin, and unverifiability against stable sources. This means Enlightenment verification tools—citation, archiving, peer review, evidentiary chains—cannot process AI output by design.
Farrell and colleagues argue large models are lossy summarizers of human-accumulated information, akin to markets and bureaucracies. They compress cultural material into usable but incomplete representations, then reorganize it—without possessing understanding of truth or falsity.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- The Human-Machine Knowledge Spiral
- What Do Large Language Models Know? Tacit Knowledge as a Potential Causal-Explanatory Structure
- Genius on Demand: The Value of Transformative Artificial Intelligence
- Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills
- Automation, AI, and the Intergenerational Transmission of Knowledge
- Large AI models are cultural and social technologies
- How to Build AI Agents by Augmenting LLMs with Codified Human Expert Domain Knowledge? A Software Engineering Framework
- Embodied LLM Agents Learn to Cooperate in Organized Teams