Neurobiological and Cybernetic AI for Manufacturing, Part 2 - with Oleg Savin of Unilever
🎯 Summary
Podcast Episode Summary: Neurobiological and Cybernetic AI for Manufacturing, Part 2 - with Oleg Savin of Unilever
This episode, the second in a series, features Oleg Savin, Enterprise Architect and MES Expert at Unilever, discussing the critical distinction between neurobiological and cybernetic AI approaches and their implications for modern manufacturing. The core narrative focuses on bridging the gap between vast data collection and meaningful data utilization through advanced AI architectures.
1. Focus Area: The discussion centers on the necessity of moving beyond current data collection practices toward true Artificial Intelligence (AI)—distinct from standard neural networks—to build the next generation of Smart Operating Systems in manufacturing. Key concepts include ontological modeling, cybernetic vs. neurobiological AI, and the convergence of IT/OT.
2. Key Technical Insights:
- Data Underutilization: Less than 10% of collected manufacturing data is meaningfully applied to generate knowledge or drive decisions, highlighting a critical need for intelligent filtering systems.
- Ontological Modeling as the Foundation: Future smart systems must be built on ontological models that allow AI to perceive, extract, and interpret high-value information fragments from massive data lakes, enabling the system to understand the manufacturing domain contextually.
- True AI vs. Neural Networks: Neural networks are useful but insufficient; true AI, as envisioned by Savin, must possess the capacity to acquire and formulate knowledge autonomously, moving beyond pattern recognition (like text generation in LLMs) to produce essence and maintain system stability.
3. Business/Investment Angle:
- Value of Ontology: The specific ontology developed for a domain is becoming the most valuable asset, representing accumulated, stable knowledge that drives algorithmic functionality.
- Eliminating Traditional Stages: The proposed framework will allow frontline personnel (engineers, managers) to autonomously design and implement smart operating modules, eliminating traditional stages of requirement formulation, programming, and installation.
- ROI through Stability: The primary function of this advanced AI is achieving high process stability by predicting and eliminating deviations, potentially rendering traditional control, accounting, and management functions redundant, leading to significant efficiency gains.
4. Notable Companies/People:
- Oleg Savin (Unilever): The expert guest, known for his insights on manufacturing conferences, driving the technical and philosophical arguments for advanced AI architectures.
- Academician Gorskov: Mentioned historically regarding early, unimplemented national economic management systems that lacked the necessary technological stack and mathematical models.
5. Future Implications:
- IT/OT Convergence: Complete convergence will occur as production staff, equipped with ontological knowledge, can configure manufacturing models directly, even in the cloud.
- Self-Organization: Manufacturing will move toward self-organization and self-management via autonomous smart modules, enabling end-to-end management of the value creation process.
- Shift in Personnel Value: Future effectiveness will depend less on IT capability and more on personnel motivation and their ability to work with information, knowledge, and models, fostering creativity.
6. Target Audience: Manufacturing Executives, Enterprise Architects, Industrial IT/OT Leaders, and AI Strategists focused on deep industrial transformation, particularly those looking to understand the theoretical underpinnings required for next-generation operational excellence beyond standard machine learning applications.
Comprehensive Summary:
Oleg Savin argues that the manufacturing sector is drowning in data (less than 10% utilized) because current systems lack the foundational intelligence required to interpret it contextually. The solution lies in developing Smart Operating Systems based on true Artificial Intelligence, which he explicitly separates from the current paradigm of neural networks. This true AI must leverage ontological modeling—a stable, domain-specific knowledge structure—to filter petabytes of data, extract relevant patterns, and maintain system stability by anticipating deviations.
Savin emphasizes that achieving this requires a fundamental shift in understanding, necessitating that business leaders acquire knowledge across four domains: philosophy (including phenomenology and epistemology), linguistics, neurobiology/cognitive psychology, and mathematics (including synergetics and fractal theory). He posits that the most advanced systems will utilize a fractal cognitive computation model of consciousness.
The practical benefits of adopting this cybernetic/neurobiological hybrid approach are profound. First, it forces the complete convergence of IT and OT, allowing frontline staff to configure complex manufacturing models directly using established ontologies. Second, it enables self-organization and self-management through autonomous modules, leading to end-to-end value chain management. Third, the primary outcome is achieving a required level of process stability for a given time horizon, potentially making traditional reactive control functions obsolete.
When advising executives on AI initiatives, Savin stresses that they must first clearly define the holistic Key Performance Indicators (KPIs) that characterize the business state (e.g., stability, efficiency, value produced, interpreted through time). They must then demand explainability from their AI teams on how these high-level indicators are formulated and performed in real-time by the underlying models. The conversation underscores that while the theory is advanced, the practical implementation of these concepts is already underway, promising a revolution in industrial efficiency driven by deep, contextual understanding rather than mere data processing.
🏢 Companies Mentioned
💬 Key Insights
"Real AI, Oleg argues, should be capable of generating knowledge, not just text"
"manufacturers are collecting massive volumes of data, but without the right AI architecture, less than 10% of it is used effectively."
"Then the third consequence, it is achieving very quite level process stability for human time horizon. In this case, the traditional functionality, like accounting, control and management, may become redundant."
"there are actually four domains of knowledge domains. It is philosophy, which includes phenomenology, epistemology, logics, there is a second domain is linguistics, it is semantics, and we have to have our own language. Then it is the science of human brain and thinking, it is neurobiology of the brain and cognitive psychology... and the fourth domain is mathematics."
"in the future, the more valuable will be the specific ontology for a specific area. A good ontology will cause much more, it is actually accumulated knowledge, knowledge."
"neural network, in spite of there being very useful for many cases, it is still not artificial intelligence, although many people tend to call it artificial intelligence."