The market for large language models such as GPT-3/ ChatGPT is just emerging. In addition to the models from OpenAI and Microsoft, the models of the US hyperscalers Google/Alphabet and Meta currently dominate. In addition, there are European solutions, such as those of the company Aleph Alpha. Likewise, the Open GPT-X initiatives from Gaia-X or LEAM from the German AI Association. Figure 14 shows the different language models according to the respective model size.
The development of these language models is characterised by extremely high development dynamics and short innovation cycles. The exciting question is to what extent German or European solutions can keep up with those from the USA and China, which are being developed with significantly higher investment volumes. An exciting approach in this context is so-called transfer learning. For example, the established language models from the USA can be used and transferred to questions posed by German companies on the basis of transfer learning.
