Prof. Dr. Volker Tresp is a professor of machine learning at the LMU in Munich, Germany, and a working group leader of the German Learning Systems Platform. He sees ChatGPT as a milestone in AI research.
ChatGPT moves the crowd. The textbot is one of the so-called large language models that are celebrated as breakthroughs in AI research.
Do these models offer real progress or are they just hype? How can language assistants be used – and what conditions do we need in Europe so that businesses and society can benefit from AI progress?
Prof. Dr. Volker Tresp from the Ludwig Maximilian University in Munich answers these questions in an interview. Tresp’s research focuses on “Machine Learning in Information Networks” and he is co-leader of the “Technological Enablers and Data Science” working group of the Learning Systems Platform.
What are large language models and what makes them special?
Volker Tresp: Large language models are AI models that use machine learning methods to analyze huge amounts of text. They use more or less the entire knowledge of the World Wide Web, its websites, social media, books and articles.
This allows them to answer complex questions, generate text, and make recommendations. Dialog or translation systems are examples of large language models, most recently of course ChatGPT.
You could say that Wikipedia or Google Assistant can do much of the same. But the new language models handle knowledge creatively, their answers resemble those of human authors, and they can solve different tasks independently.
They can be extended to handle arbitrarily large amounts of data and are much more flexible than previous language models. Large-scale language models have moved from research to practice within a few years, and of course there are still shortcomings that the world’s best minds are working on.
But even if the systems still occasionally give the wrong answers or fail to understand questions correctly, the technical successes that have been achieved are phenomenal. AI research has reached a major milestone on the road to true artificial intelligence.
But let’s be clear: the technology we’re talking about is not a vision of the future, it’s already here. Anyone can use those language assistants and chatbots through a web browser. Today’s language models are real game changers. In the next few years, they will significantly change the way society, science, and business deal with information and knowledge.
What applications can be built with large language models – and what are the requirements?
Volker Tresp: Language models can be used for a variety of applications. They can improve information systems and search engines. For service engineers, for example, a language model could analyze thousands of error reports and problem messages from previous cases.
For doctors, it can help with diagnosis and treatment. Language models belong to the family of generative transformer models, which can generate not only text, but also images or video. Transformer models create code, control robots, and predict molecular structures in biomedical research.
In sensitive areas, of course, it will always be necessary for humans to review the results of the language model and ultimately make a decision. The answers given by language models are still not always correct or off-topic. How can this be improved? How can we integrate more sources of information? How can we prevent language models from incorporating biases in the underlying texts into their answers?
These are fundamental questions that require a lot of research. So there is a lot of work to be done. We need to nurture talent in AI, and create professorships and research positions to address these challenges.
In addition, if we want to use language models for applications in and from Europe, we need European language models that can handle the languages here, taking into account the needs of our companies and the ethical requirements of our society. Currently, language models are created – and controlled – by American and Chinese tech giants.
Do large language models only benefit large companies?
Volker Tresp: Small and medium-sized companies can also use language models in their applications because they can be adapted very well to individual company problems. Of course, mid-sized companies also need technical support.
Service providers, on the other hand, can incorporate the customization of language models into their business model. There are no limits to the creativity of companies when it comes to developing solutions.
Similar to search engines, use cases will multiply like an avalanche. However, in order to ensure that small and medium-sized companies do not face financial hurdles, we need large foundational language models under European leadership that allow free or low-cost access to the technology.