Research on artificial intelligence (AI) began with the desire to understand whether and to what extent computers can solve problems in a way similar to the human mind. Many of the early works related to AI focused on creating mathematical models and algorithms capable of performing tasks requiring intelligence, such as reasoning, learning, perception, or using language.

The development of large language models is currently one of the key topics in the field of new technologies, which is why we encourage you to read further parts of this article, which include comments from the CEO of Inero Software, Andrzej Chybicki, on the evolution of this issue over the past years.

Natural Language Processing. A New Approach to Data Analysis

Techniques for processing speech, text, and images have been used in industry and science for many years. They formed the basis of products based on so-called NLP (Natural Language Processing) models. These are algorithms that could understand individual words, later map relationships between words, and combine them into sequences. They began to understand and distinguish the meaning of the same words in different sequences (in other words, they distinguished context) in order to ultimately be able to interpret the meaning of entire sentences. To say that NLP models “understand” language is somewhat of an exaggeration; nevertheless, current NLP language models are able to assign individual sentences to so-called intents (intentions), i.e., distinguish their meaning. Then, by creating appropriate software, we are able to handle simple conversations, dialogues, or respond to simple questions – this is how chatbots have worked until now.

To illustrate this number, we might compare it to reading a several-hundred-page book in a few seconds.

Words Grow in Power: Large Language Models

LLMs, or Large Language Models, which have been developed over the past few years, represent a technological breakthrough because they no longer function merely as classifiers of user intent, but by using trained sets of input and output similarities, they are able to find connections between longer sequences of words (tokens). Large language models are now capable not only of analyzing individual sentences but also significantly longer texts. This is referred to as the length of context (or context window), which currently can be as much as 128,000 tokens. To illustrate this number, we might compare it to reading a several-hundred-page book in a few seconds.

Level of Understanding: Logic in Many Ways Similar to Human Logic

Students sometimes ask, “do computers understand text?” The answer to this question is difficult, if we do not define what we mean by “understanding”. Creators of language models and researchers deal with creating various measures, criteria to be able to evaluate whether a model “understands” a given problem. Nevertheless, in the simplest approach, we can say that a model understands a given issue if the responses (sequences of words), which it generates, are consistent with how a human expert in the field would respond.

I believe that in the context of AI technology development, the difference between how humans communicate (speech, writing) and computers (protocols, databases) is gradually blurring. It seems that in the near future, AI models will support humans in an increasing number of everyday tasks. They will be able to automatically retrieve information to send a response to a customer, suggest to employees the answers to questions, or handle communication through messengers.

Models will be able to browse a whole range of information, including documents and instructions covering, for example, orders, emails, contracts, or invoices, then infer based on them, and subsequently provide the employee with ready data needed for a report, an offer, or other operational actions.

Virtual Assistants: Will Employees Gain Real Help?

Much indicates that AI assistance in work is the future for many areas, yet I still see a big problem. Despite the fact that digitization, mainly during the pandemic, has accelerated rapidly, many companies and institutions still process a large part of information in an unstructured form, i.e., in the form of documents, contracts, instructions, and email messages. The amount of information that employees deal with daily is enormous.

Ultimately, models will be able to browse a whole range of information, including documents and instructions covering, for example, orders, emails, contracts, or invoices, then infer based on them, and subsequently provide the employee with ready data needed for a report, an offer, or other operational actions. Furthermore, a potential area of application will also include validation and checking of documents, analysis of dependencies between two different documents, or several sources of knowledge.

Virtual Assistant Speeds Up Daily Work

An example might be a situation where a bank employee receives a credit application along with attachments such as account status confirmation, employment confirmation, tax statement, and others. Currently, employees often have to manually check whether the data contained in such documents are consistent, e.g., whether the number given in the credit application matches the one on the employment certificate. Whether the bank account number written in the application is correct, and whether the dates of the issued income certificate are current, among many other things. Similarly, in the insurance sector, where issuing a policy requires manually checking VIN numbers on documents, registration certificates, and comparing them with data about inspections, policies, etc.

Implementing tools based on the LLM model into organizations can bring measurable benefits in the form of automation and increasing work efficiency. It is a mistake to claim that software can replace an employee. It can help them. Analyzing the European market, we see demographic trends resulting from an aging society and the decreasing availability of skilled workers.

In summary. Natural Language Processing (NLP/LLM) models increasingly support people in everyday tasks through automating responses to customer inquiries, assisting employees, and managing communication through platforms. These technologies not only increase work efficiency through the automation of routine tasks but also enable advanced operations like document analysis or data integration, which translates into broad applications across various industries and professional levels.

Andrzej Chybicki, PhD Eng.
CEO, Inero Software Sp. z o. o.
andy@inero-software.com

Artykuł Andrzej Chybicki: Models will be able to instantly review an entire repository and then provide the employee with ready-to-use data pochodzi z serwisu Inero Software - Software Consulting.

At Interizon Day 2023 Waldemar Kołub co- founder of Inero Software presented a success story about using Neuronal Networks to optimize the production process.

Farm Frites Poland is leading producer of potato products, supplying products to, among others, one of the largest fast-food restaurants in the world. The company has been operating for over 30 years, and its headquarters is located in Lębork. The modern machinery park consists of three production lines.It employs over 270 people, and the quantity of potatoes used for production over a year exceeds 250,000 tons.

“The problem our client had was to estimate the production yield based on the parameters of the raw input material, so how much of the final product we can obtain from the given amount of the source material, or, to put it in layman’s terms, how much fries we can get from a potato. And it turns out that it is not an easy question to answer.” –  Waldemar Korłub mentioned at his presentation.

Challenges for the development od the company was:

DATA OF THE RAW MATERIAL:

• Contracting for raw material supplies is planned several months in advance and constitutes a significant element of the enterprise’s manufacturing process and pricing policy.
• Estimation of yield, i.e., the difference between the weight of raw material and the final product expressed in percentages (EPY), allows for adjusting the raw material price to the expected result.

PRODUCTION OPTIMIZATION:

• Identification of raw material characteristics that most significantly affect the yield (EPY)?
• Indicate what yield can be expected based on the given raw material?

“So there are a lot of parameters that describe the source material. All in all, there are over 40 different physical traits that need to be taken into account that might influence the final. Yield with the set of parameters. We perform exploratory data analysis to find correlations between the parameters and the final results. By doing that, we are able to identify the parameters which influence the final results the most. Also, we looked for correlations between the parameters which allowed for dimensionality reduction in the problem space and, as a result, it allows for speeding up the machine learning process.”

Machine learning is method used for “feeding” machines with data so that based on this data and their own errors, they learn the rules to solve a given problem. It can be assumed that this is a role reversal, where previously, by defining the rules and utilizing the data, we sought answers.

“With a well-defined set of parameters, we could move on to model research. We’ve tested several different machine learning approaches, including regression models, decision trees, support vector machines, and neural networks. In the end, it turned out that the best results were achieved with neural networks, with the mean error at around 1.2% for the validation data.“

LEARNING

Learning is based on providing training examples (historical data) in the form of:

• Average raw material per run: Green, Rusty, OWG, Color (actual values)
• Average yield per run: EPY (e.g., 45%)

The network minimizes the error between the actual value of EPY (45%) and what it predicts itself (e.g., 54%). The metric used is MSE (Mean Squared Error).

IMPLEMENTATION AND VALIDATION

The model was implemented in the client’s infrastructure. Thanks to the introduction of the solution, it was possible to minimize the forecast error to just 3%. This means that accurate estimates can already be made at the stage of contracting raw material supplies.

“This preliminary model research was conducted on a subset of the dataset, and the model that was decided to be the best was later on deployed in our client’s infrastructure and we performed the full training using the complete set of production data from historical production runs. The final model was deployed to production with the mean squared error below 3% for the real-world production runs.”

This solution was developed according to our build, operate, transfer model. “We build the solution using our knowledge and expertise in the first step. Then we helped with setting up the infrastructure and the processes to meet the operational requirements, and finally, the product was transferred to the client, who receives the full ownership of the solution. With this process, we are not only trying to create solutions for individual problems but rather to create a platform for the ongoing development of innovations in the environment of our clients.”

At Inero Software, we believe that software goes beyond just the technical; it’s also about people, business, and delivering on sustainable long-term objectives. By leveraging the BOT model, we establish a lasting innovation framework within our client’s operations, enabling them to set and achieve long-term business goals through the power of innovation.

Waldemar Korłub, Eng. PhD

CO-FOUNDER INERO SOFTWARE

waldemar.korlub@inero.pl

Artykuł Case study: Deploying an AI-based solution at Farm Frites Poland pochodzi z serwisu Inero Software - Software Consulting.