AI Infrastructure
/
AI Infrastructure
In this project, the Howest AI Lab and the Howest Energy Lab make a joint effort to expand their existing AI research infrastructure, with the aim of being able to develop and train the new generation of AI models.
Main section
Quick facts
/
Expansion AI research infrastructure generative AI.
/
Training new generation AI models
The emergence of generative AI, such as Transformers and Diffusion models, has led to a greater demand for powerful graphics computing power (GPU) and associated video memory (vRAM). The current AI infrastructure of the labs no longer meets these requirements and also does not provide the capacity needed to serve the growing number of researchers. The new infrastructure will solve this shortage of computing power and will enable researchers to explore new AI-related topics, which in turn will lead to knowledge transfer to education and the field.
Within Howest, there is no research infrastructure yet that is capable of researching Generative AI in an efficient way. Alternatives such as Cloud infrastructure or the Flemish Supercomputer Center are either less cost-effective or less flexible for the purposes of the labs. This makes the new infrastructure of great value for both the university of applied sciences and the companies with which the research labs collaborate. Two interaction moments for knowledge transfer are planned.