The topic of AI is currently on everyone’s lips. How do you assess the current discussion?
Generative artificial intelligence (AI) is undoubtedly one of the groundbreaking innovations of our time. In just a few years, it has gone from pure hype to an enormously powerful tool that offers considerable advantages for companies in all industries worldwide. Thanks to ChatGPT and many other publications, users can now generate media content, answer customer queries in real time and automatically generate program codes, among other things. AI solutions, and ChatGPT in particular, are also opening prospects for logistics that we would hardly have dared to dream of just a few years ago – and the potential is far from exhausted.
Where do you see the greatest benefit of ChatGPT?
ChatGPT is a great tool for creating written content and answering questions about general world knowledge. Thanks to major advances in research, unstructured data such as images, text, speech, and audio can now be automatically processed and analyzed via this application. One of the greatest benefits however lies in its accessibility. The ease of access and general usability enables everyone to directly interact with artificial intelligence in their everyday life – not only improving their AI literacy, but also teaching the limitations of AI and large language models in general.
Can you explain this in more detail?
The priority of ChatGPT and other applications is to understand and produce language that is grammatically and semantically correct. In order to be able to give a coherently sounding answer, large language models tend to hallucinate and deliver factually incorrect answers when they are lacking the proper information. It is therefore important for companies to continuously evaluate the results of the model on a professional level and ensure that they are fit for purpose. However, this requires specific expertise, which we at Lufthansa Industry Solutions are happy to provide.
What does this support look like?
In general, we closely monitor the developments on the market and regularly benchmark available large language models against each other. Further, one of the main trends that emerged from the ChatGPT hype is called retrieval augmented generation. Picture a handling agent, that has a process specific question. While ChatGPT might not be able to answer that question per se, we can supply the required process documentation (such as handling manuals) to the model and impart that knowledge, enabling a semantic search on several thousand pages of documents, within a few seconds.