Manager(s) |
Stefan Decker |
Funding | BMBF |
Project Start | November 01, 2022 |
Project End | December 31, 2025 |
Status | Running |
Website | https://westai.de |
Recent advances in AI technology have opened up enormous potential for the automation of intelligent tasks
previously only accessible to humans, with corresponding opportunities for doing business more effectively,
creating entirely new products and services, and achieving advances in all areas of science. Consequently,
the demand for advanced AI technologies and services is increasing worldwide and the effective use of AI is
considered central to the competitiveness of companies and entire economies. While AI uptake has been
quick in other regions of the world, there is a considerable gap in AI adoption in Germany in both companies
and science due to the lack of access to hardware resources and adequate services to support users, along
with a corresponding research gap in the development of scalable methods and systems.
At the same time, whereas the current generation of deep learning AI systems is becoming more and more
available, a fundamental and disruptive change in AI is about to change the game once again: the switch
to the use of very large pre–trained models that complement task–specific training to result in a quantum
leap in predictive accuracy, quality and range of tasks that can be addressed. As shown for example by GPT–
3 and other transformer models, using large models appropriately allows to tackle tasks such as understand-
ing and even generating text and images with unprecedented potential. Yet at the same time, these models
require computing resources of the highest order of magnitude and are still difficult to train and to transfer to
new domains – in particular for smaller companies and organizations.
Thus, there is a clear need to both foster AI adoption in economy and science at large by offering computing
resources and services, while at the same time driving forward the next generation of large AI models with
corresponding research, better hardware and novel offerings. The proposed WestAI service center therefore
combines a strong orientation on provisioning best–of–class scalable hardware with a clear and unique
research focus on transferable large models, making them better trainable and transferable to new domains
especially for smaller players. Driven by these technological advances and its best–of–class hardware offer-
ings, WestAI offers support to increase AI uptake with low–barrier, high–quality services, training and con-
sulting.