Artificial Intelligence (AI): from a prompt it analyzes data and interprets them for facial recognition, machine learning, language processing and translation, virtual assistance, and automation. Generative Artificial Intelligence (GenAI) is an evolution: starting from data analysis, it can write content, code, and designs, and then generate new results that did not exist before. In short: AI observes to decide, while GenAI imagines and builds.
by N. S.
…what are the implications for an industry as highly automated as packaging? To understand this, we met Massimo Monguzzi, Head of Research and Development at Cama Group, an international reference point for secondary packaging.
From code to its functions, the evolution of writing
“From an initial curiosity, AI and GenAI have now become work tools. The first area has been the writing of software code, which is becoming increasingly high level. We are rethinking the approach to programming to automate repetitive operations such as code writing,” begins Massimo Monguzzi; “the competence of the software developer, a key role for a company like Cama, must understand its ends. On the one hand there will be less time spent writing code, with more attention to functionalities. Delegating code generation to AI allows developers to focus on higher value-added activities. Writing code will be like tightening a screw: a necessary activity, but not strategic.”
Technical documentation is a job for AI
A second area is technical documentation: manuals, technical files, risk analyses, declarations of conformity. Complex, regulated productions, costly in time and resources.
“GenAI speeds up the creation of these materials if it is integrated in a data-rich context from which it can learn. Cama has a significant documentary asset made up of thousands of drafted technical files. In similar scenarios GenAI proves effective, even if it cannot replace human control, but it increases productivity. It also reduces errors and enables greater standardization of technical content.”
Training GenAI: specific training for every context
An underestimated feature the training of AI and GenAI to the specific context in which they will have to perform tasks. One belief is that they can operate autonomously. The opposite is true.
An AI model must be trained, and this means immersing it in a real technical environment, as is done with a new human resource. In this way AI ceases to be a generalist system, and he learns specialist skills, as in the case of packaging.
“Each sector has its own language with its terminologies. In packaging the term pocket has one meaning, which changes in fashion or, for example, in mechanical engineering. If the AI does not know the correct term to use for each context, it makes mistakes.”
In manufacturing, every machine, every plant, and every project are unique. GenAI must be trained with specialist knowledge tied to the production context, and this is an important internal task for the company.
Ever more advanced but user-friendly machines
The complexity of the machines becomes their limit when operators cannot be found who can use them at their best. To overcome this obstacle, Cama has long been developing user-friendly interfaces.
“For controlling a machine that may be very complex, the customer expects a simple interface. Here software is important because it allows our machines to be used even by operators with few technical skills. In fact, a widespread problem in many countries is frequent staff turnover and there is no time to train them. Having a user-friendly machine is related to staff experience in a balanced way the staff has. GenAI also helps in writing the software specific to operator interfaces.”
The help also applies to students fresh out of university or technical institutes; with the support of AI they can become operational in a short time, making the labor market more dynamic.
We are in the middle of a cultural change that redraws the boundaries between humans, machines and artificial intelligence. If the packaging industry requires specific skills to be 100% operational. AI and GenAI shorten the time of this training.
“If on the one hand today’s machines are more complex than in the past, fortunately they are also more user-friendly. The path taken is to make sophisticated systems simple according to the Industry 5.0 paradigm, which brings humans and machines closer together through interfaces, contextualized language and shared goals.”
GenAI and data correlation
Cama’s Research and Development division is developing algorithms to study machine data in depth. Their analysis, machine learning, and deep learning contribute to process optimization by identifying correlations between operating parameters and performance.
“Production efficiency is also correlated with environmental variables. Humidity, for example, varies according to the seasons but also to the geographical area and affects the machinability of materials and the quality of packaging. The same goes for temperature, which accelerates the wear of various parts. Knowing in depth the functions of these parameters, it makes possible to improve the setting of the machines and prevent potential problems.”
Can Artificial Intelligence make decisions?
Computer vision for pick-and-place must make decisions quickly. “Machines must think and decide in real time which product should be picked and at what angle to place it, discarding out-of-spec parts. The actions must be very fast as well as very repetitive—perfect for an intelligent algorithm. This logic can be extended to the micro-decisions of the machines’ operating sequences, with the aim of improving their overall efficiency. It is not a matter of overturning the machine’s behavior, but of optimizing it in a continuous and adaptive way. GenAI also makes the difference in quality control: if we think of the quality of a biscuit, an intelligent system distinguishes a burnt product from a perfect one. It works with the regular shapes of biscuits but also with irregular products such as croissants”, explains Massimo Monguzzi.
The same principle applies to OCR reading (optical character recognition), an application traditionally difficult due to variable printing conditions, now more reliable with AI and widely employed in quality control of the labeling process.
What is Edge computing? The future is the machine that reasons
Soon, packaging machines will have a small “brain” to analyze in real time the data generated by on-board sensors: Edge computing. In this way it will be possible to know the machine’s work history, detect anomalies in milliseconds, optimize performance, and reduce stoppages and waste, producing perfect packs ready for the shelves.
Edge computing on the machine, together with AI and GenAI, closes the loop: from the language for software to final quality control throughout the production process. An extremely advanced system that learns by working. The result?
- Superior performance: the machine self-optimizes instantly before the pack leaves the packing station for the shelves.
- Unprecedented flexibility: generative systems adapt formats, recipes, and motion sequences as quickly as market demands change.
- High ease of use: user-friendly interfaces and automatic suggestions towards a natural human–machine dialogue.
Even with the most technological tools, evolution does not stop to reach the goals of packaging ever more effectively: efficiency, flexibility, and sustainability.
