The AI-based tool ChatGPT from the company OpenAI is a source of immense fascination worldwide. Almost daily, the media reports on amazing features, but also on bugs and weaknesses of the large language model. So we ask ourselves the question: Is ChatGPT suitable for industrial use?
What is ChatGPT?
In a few words, ChatGPT is a transformer-based large language model. It has been trained on a large amount of data from the internet to generate texts on its own. ChatGPT generates text by making predictions about the next section of text based on probability calculations. Because of the way it works, ChatGPT can also be described as a prediction model.
The training method used is unsupervised learning – a subset of machine learning that works without human supervision or input.
General features and functions
Human-like creativity with ChatGPT
ChatGPT is best known for its ability to generate text and answer questions. With just a few prompts, the model composes emails, statements, summaries, tweets or checklists of questions. ChatGPT can also write programming code.
It is particularly useful as a stimulus for brainstorming or for compiling research. In this way alone, the tool can make everyday work easier for all knowledge workers. The capabilities of ChatGPT reach a level of creativity that comes very close to that of humans. No wonder, because creativity is all about gathering inspiration, information and knowledge from many different sources and combining them in new ways. This is made possible by the huge treasure trove of data in the background of ChatGPT.
Critical handling required
Collaboration between the human and the chatbot is still essential for an optimal result. In most cases, multiple inputs are required to achieve the desired result. The more precisely you ‘prompt’ – that is, enter commands and describe your request – the better the result. However, good prompting is not easy and requires expertise and practice.
There are also some drawbacks to ChatGPT that need to be considered before using it. For example, the AI model still struggles with nuance and judging what information is important or less relevant. Sometimes it gives completely wrong information that sounds reasonable and plausible. Questions that it cannot answer out of ignorance are often dismissed with made-up answers that have no basis in reality. So caution is advised here.
ChatGPT cannot think critically. Therefore, users should never blindly rely on the information provided. It is essential to check the information to ensure that it is correct and properly categorised. Especially when making decisions and dealing with sensitive issues, it is essential to do your own research and form your own opinion on the subject. As mentioned above, it is important to remember that the knowledge base of the tool ends at a certain point in time and may not be completely up to date.
Legal protection
There are also technical and legal issues that need to be addressed before deploying ChatGPT in an organisation. It is often a challenge for companies to integrate ChatGPT into existing systems and processes. These include difficulties in integrating the necessary data or scaling up if companies need to handle a particularly large number of enquiries. In addition, ChatGPT is a generic model and in some cases needs to be trained for specific terminology, corporate policies or other requirements.
In addition, data protection must always be guaranteed when using it, especially with regard to Europe and Germany. Many questions remain unanswered. However, the first mistakes should warn corporate users to be careful. Under no circumstances should sensitive data be entered into the system. Ideally, companies should draw up a policy before using the system, explaining exactly how and in what cases it may be used.
This includes protection against copyright infringement. Initially, headlines about stolen code made users nervous. This issue is also highly sensitive with similar image creation tools. Guidelines should therefore also ensure that copyrights are not unknowingly infringed.
Corporate use
Once these questions have been answered, there are many ways that businesses can use ChatGPT.
Increase customer satisfaction
Obvious applications include automating communication between businesses and customers. The chatbot can quickly and confidently respond to enquiries, explain issues or handle complaints.
Businesses can also use the ability to quickly analyse large volumes of text, for example to perform analysis on large volumes of unstructured data such as emails, social media posts or customer feedback. By analysing this data more quickly, it is possible to identify complaints more quickly and thus improve customer satisfaction.
To further improve customer satisfaction, it is also possible to use large language models to make personalised recommendations based on user behaviour or customer communications.
Optimise processes
Businesses can also use ChatGPT to improve the efficiency of day-to-day processes. For example, it can be used to write product descriptions or prepare translations. The chatbot is also useful for initiating research or identifying sources.
Long texts, such as contracts, can be analysed in no time at all to display the most important information. This saves employees time. It also ensures that details such as deadlines are not overlooked, which reduces misunderstandings and errors and leads to a smoother workflow.
Programmers benefit from ChatGPT in many ways. The AI model not only outputs finished code, but also provides assistance with code problems. For example, it suggests alternative solutions or tools. So even if you don’t use ChatGPT to get a finished result, it can be a valuable companion on the way to a solution.
Ready for industrial use?
At the moment, the chatbot itself cannot control robots. However, it is important to remember that it is based on predictive AI algorithms that can control robot movements. Just as ChatGPT can calculate the likely next word from a large amount of text data, similarly structured AI models can make decisions for navigation and interaction with robots based on image or sound data – for example, by calculating the likely next movement of a robot to pick up an object or turn a corner. OpenAI is currently working on a robot that can mimic human movements and behaviours, and thus learn to perform human-like activities.
ChatGPT can already be used in industry. For example, production workers can enter the facts of a problem in a free-text mask and then receive troubleshooting instructions. The AI model can also provide a ticket assignment for maintenance tasks, supporting the technical service. It tracks the exact mood of the person making the request and adapts accordingly. In this way, the tool creates a better user experience.
First steps – where to start?
When a company decides to use ChatGPT, it is advisable for several people to test the tool on a small scale to familiarise themselves with it. Before planning a large-scale project, key stakeholders should get together to analyse how the AI model can be integrated into the company’s own processes. AI for AI’s sake is not the answer. The main objective should be to derive real value from the tool.
Stakeholders should not only include members of the IT department. As mentioned earlier, legal issues should be covered. Data protection officers or legal advisors should also be part of the decision-making body.
The start of a big change
Much more than a media phenomenon, ChatGPT can be a helpful companion, source of ideas and sparring partner in many facets of the working world. Whether used to facilitate everyday communication, optimise customer service processes or even control robots, ChatGPT and a host of related AI tools will change the way we work forever.
Images: Adobe Stock
Deepen your knowledge with more magazine articles on the exciting topic of artificial intelligence and stay at the forefront of technological developments:
