Machine learning and Artificial intelligence aren't new. But the release of ChatGPT sparked a hype throughout the tech scene in recent years. The transformers architecture behind ChatGPT is grounded on work published by Vasnawi et al. (2017) from the Google Brain Team entitled "Attention is all you need".

The advancement of this architecture actually lies in a reduction of model complexity (as compared to traditional convolutional and recurrent neural networks), which allows significantly better parallelizability and hence its much more efficient means of training and use.

Unfortunately, OpenAI decided to rethink its Open Source approach and replaced it with a SaaS model. A huge and vibrant Open Source scene emerged quite recently to counteract this move and produce alternatives that can run on your own hardware, to ensure privacy and long-term stability without having to rely on a single large company.

Within this presentation we'd like to introduce the audience to the Open Source research and programs for large language models (LLMs), their potential fields of application with regards to the field of serious business applications and show some experiments that we've done internally to evaluate its potential.

The session will be a condensed bit of theory and introduction to practical sources of information for developers, the presentation of a few experiments we did and hopefully an interaction with the OCA community on where and how the presented Open Source tool chain may be of use.