Energy Efficient AI

Purpose and goal: ** Denna text är maskinöversatt **

The goal of the project is to develop computational (and thus energy efficient) algorithms for training AI models. It is a fact that this will be unsustainable in the long run, as there is not enough energy to serve everyone who would want to implement AI. We have developed new algorithms and the results will be published through at least one conference paper within one of the AI field´s top conferences.

We will take the collaboration to the next step and build an open-source community similar to HuggingFace. Expected results and effects:

The project resulted in in-depth collaboration which, in addition to Prof. Stephen Boyd at Stanford, includes Prof. Mert Pilanci and Prof.

Jure Leskovic at Stanford. We will take the collaboration to the next step and build an open-source community similar to HuggingFace. We have developed new algorithms and the results will be published through conference papers in the AI field´s top conferences.

We will also apply for a patent to secure the IP. The expected effects are that we get a global open-source community and a product for many companies and organizations. Approach and implementation:

The applicant conducted a longer visit to Stanford and met with Professors and started working on the algorithm development with the professors and their students. We started off by formalising the algorithms, then started implementing and testing the algorithms on both simulations and open-source data. Successful tests confirmed the superiority of the the new algorithms. We started the process of writing a couple of scientific papers to disseminate the results to the public.