Modern manufacturing stands to benefit from integrating AI. From improving efficiency and productivity by automating repetitive tasks to reducing unplanned downtime and cutting down on repair costs through predictive maintenance, the potential benefits are numerous.

However, integrating AI in manufacturing is not without challenges. A 2019 article for the journal Engineering from the Chinese Academy of Engineering and Higher Education Press cites complexity and uncertainty as significant challenges in manufacturing in the coming years. Overcoming these challenges in short order does not secure the role of AI in manufacturing, as resistance to replacing humans remains.

A later 2021 paper published by Harvard Business Review cautions against AI replacing human intelligence by emphasizing the expansive aspect of human abilities and the fact that they don’t require a steady supply of external data, the way that AI does.

Emerj Senior Editor Matthew DeMello previously sat down with MES Expert from Unilever, Oleg Savin on the ‘AI in Business Podcast’ to talk about the challenges and potential of AI in manufacturing.

The following article will focus on two key takeaways from their conversation:

• Outlining three key challenges for next-generation AI systems in manufacturing: Recognizing the role of autonomous operations and data as a valuable resource in decision-making and knowledge accumulation.
• Understanding thinking in how the human mind works: Differentiating between the neurobiological and cybernetic approaches to knowledge to define intelligence within a manufacturing context.

Listen to the full episode below:

Guest: Oleg Savin, Former MES Expert, Unilever

Expertise: Engineering, Manufacturing, Data Science

Recognition: Oleg Savin spent over a decade at Unilever, providing engineering expertise and leading advanced technology adoptions across several business divisions. Before his retirement in 2024, Oleg offered his thoughts on the distinctions between Neurobiological and Cybernetic AI for engineering conferences worldwide. 

Understanding Thinking in How the Human Mind Works

At the beginning of the podcast, Savin starts by responding when asked about what he sees as the most significant problems currently facing the manufacturing space, given the current state of AI adoption in this sector.

Savin says that the core functions of human intelligence are characterized by three core abilities:

• To anticipate changes in the environment
• Adapt to changes or change the environment
• Try to ensure the stability of that environment

He goes on to explain that artificial intelligence in the manufacturing space should perform the same intelligent functions, including:

• Predict the state of the manufacturing environment
• Adapt the manufacturing environment to anticipated changes by proper operation
• Take measures to ensure the sustainability of the manufacturing environment

Savin also explains two approaches to understanding how thinking in the human mind works. He first explains the neurobiological approach and how intelligence is considered a hierarchical neural network structure, emphasizing that the visible neural network is only the base layer, or connectome, but other structures are needed to interact with the environment.

Savin mentions that intelligence is defined as the ability of a system to achieve its goal in a range of environmental conditions. He explains that this is a universal mechanism that nature has developed in the evolution of biological systems.

He contrasted this with the cybernetic approach to knowledge, which equates intelligence with the process of thinking as logic and knowledge, where the knowledge representation model is a semantic network. Within this topology, intelligence is defined as problem-solving technologies based on the idea of using domain knowledge. 

Additionally, Savin explains two major approaches toward defining intelligence so that we can understand what intelligence is. 

He says that artificial intelligence should mimic human intelligence and its imminent functional characteristics. 

• Availability of knowledge: To operate effectively, AI requires domain-specific knowledge in the form of an ontology and a set of global axiomatic logical rules.
• Behavior adaptation and goal setting: An AI system should be capable of adapting its behavior at discrete intervals in a dynamic, changing environment.
• Holistic perception: An AI system, given a fragment of reality, should be able to perceive a complete “image.”
• Minimization of Discreteness through Recurrent Feedback: AI should minimize the fragmentation of its perception by using recurrent feedback mechanisms.

Outlining Three Key Challenges for Next-Generation AI Systems in Manufacturing

Savin explains three key challenges that next-generation automated systems in manufacturing should address:

• Smart operations and autonomy: AI should help make manufacturing operations self-sustaining and stable.
• End-to-end integration of supply chains: AI should make it easier to integrate supply chains end-to-end and enable interoperability across systems, potentially through blockchain technology.
• Data as a valuable resource: Savin calls for a shift in mindset where data is viewed as a valuable resource and used to accumulate knowledge patterns and enable effective decision-making.

When asked to explain the differences between automation and autonomization in how AI systems will operate in practice in a manufacturing context, Savin provides unique insight. Savin mentions that autonomy and self-management in business, especially in manufacturing, align with the broader goal of reducing costs. Savin mentions two trends where this is seen: robotization and autonomous vehicles. He mentions Industry 4.0, the current phase of the Industrial Revolution, characterized by the use of digital technology to improve manufacturing and supply chain operations.

Savin explains that in this context, cyber-physical systems have their own “brains,” which allow them to operate and function autonomously. He mentions that the trend in manufacturing is toward autonomous systems while acknowledging that there are social risks, such as the potential elimination of jobs.

He explains how, in different production stages, the systems will communicate with each other based on a common ontology or shared understanding, and human involvement will be limited to supervising and making decisions only when prompted by the system.