Using AI - but the right way

A brief look back at a 100-billion-dollar debacle: In February 2023, the chatbot AI combination ChatGPT, developed by the US company OpenAI, broke all user records. It uses artificial intelligence (AI) to create fluently readable texts within seconds, from school essays to instruction manuals. Microsoft announced in February that it would integrate the intelligent chat solution into its Office package and the Bing search engine. This was the first time in years that Google faced a development that could threaten its market power in the search engine sector (share in Germany: 98 percent).

Google countered with an alternative called "Bard". But in the rush, the Google team made a painful mistake: In the commercial, of all things, "Bard" gave the wrong answer to a user question. As a result, the share price of the parent company Alphabet dropped by nine percent within a day - the stock market value fell by 100 billion dollars due to the wrong answer.

This episode shows us two things: The integration of AI in chatbots is causing excitement like no other topic in IT. At the same time, Google, of all companies, has demonstrated: In practice, the new digital world is still far from perfect. In its latest version 4.0, ChatGPT recognizes images and can program. But its database only extends to the year 2021 (as of March 2023). Nevertheless, a machine that knows intelligent answers to all questions and can formulate them perfectly or output them as speech opens up an infinite number of possible applications. The Würth companies can also benefit from this.

 

AI in the IT of the Würth Group

The IT department of the Würth Group has been researching and developing AI for several years.

Why is the whole world talking about AI solutions with a chat function? To anyone entering the subject, it seems almost unbelievable that a system can behave so humanly that the differences become blurred. Nowadays, a chatbot can not only generate answers from a database, but can also be programmed to learn by asking users questions.

In addition, there is the example of ChatGPT. We are dealing here with a huge leap in generative language models. The underlying language model GPT (abbreviation for Generative Pre-trained Transformer) has been trained with huge amounts of data. It is a Large Language Model (LLM), which in version GPT 3.5 has over 175 billion parameters and 800 gigabytes of storage capacity. The result is text that is hardly or no longer distinguishable from human-generated text.

 

Resources are freed up

LLMs will completely change the way we work. There are many positive effects, such as the more natural conversational behavior of chatbots, such as those used in service and support. LLMs will also make it fundamentally easier to disseminate knowledge. In general, automation creates free resources, especially in companies, which can be used productively.

On the other hand, the use of generative language models also involves risks. The information output does not always correspond to the facts. Statistical evaluations and sophisticated pattern recognition that accesses huge data sets are not yet a substitute for common sense. In addition, the IT of the Würth Group offers its customers reliable standards, for example in the area of security. This minimizes the risk involved in working with software-based systems. The use of AI systems, on the other hand, can have previously unforeseen consequences, for example if the underlying data changes unexpectedly. Therefore, the fundamental question of responsibility arises in the course of AI development: Who is liable if the AI works incorrectly and causes damage as a result? For this reason, the main task of the IT of the Würth Group is to make the benefits of LLMs available to the Würth Group without risk and to make the data available to the business units.

 

Look into the future

Whether production, sales or service: AI or ML will find its way into all areas where processes can be digitized and optimized on the basis of these technologies. In 2023, Würth launched development projects in logistics, for social media channels and for bills of quantities in the management of construction projects, among others.

 

Automated processes

In particular, automated, rapid pattern recognition from data will enable faster, automated and ultimately better decisions. In many cases, humans will no longer make decisions, as processes will be holistically automated. AI and ML should only be seen as partial steps here. "What is important is a holistic and process-oriented end-to-end view, i.e. a view of complete processes and their optimization," says Stefan Beer, Head of Business Intelligence at Würth IT. If, for example, an ML algorithm results in a product recommendation, this can be submitted to the customer via various communication channels without the sales representative having to take action himself.

What is certain is that AI is becoming enormously important in all areas of life. The relevant technologies can drive economic growth, as can scientific advances that benefit everyone. Whether trade, health, transport or the environment, the possibilities are limitless. The example based on the IT of the Würth Group shows what it takes to make good use of them: the necessary know-how and a clear focus on things that create added value - in the case of the Würth Group, for its employees and, in particular, for its customers.