The Ministry of Digital presented the Polish language model PLLuM and its improvement plan. From now on, the model is available to everyone.
PLLuM (Polish Large Language Model) is simply a household of artificial intelligence models that allows to process and make texts in Polish. Models developed by Polish experts and experts in the field of IT and linguistics will support the improvement of digital competences and innovation in public administration and business. The announcement of the start of work on this model took place in December 2023:
PLLuM Uprising – Polish open large language model
– PLLuM is proof that we can make modern technologies on our own terms, in our own language, for the benefit of citizens and citizens. We make the foundation for intelligent public services and innovation, which will be a real support for both administration and business – says Deputy Prime Minister and Minister of Digitization Krzysztof Gawkowski.
Depending on the selected variant, the PLLuM models are available in 8 to 70 billion parameters (for comparison GPT-3.5 is simply a model with 175 billion parameters, and the current GPT-4 has 100 trillion parameters). PLLuM is simply a flexible and scalable model, smaller models according to the ministry work in fast tasks, while larger ones offer higher precision and contextual consistency in the meaning of Polish.
The PLLuM household includes models in MoE architecture with a balanced choice of experts and specialist RAG models.
PLLuM developers emphasize that their group of models is based on ethical data acquisition – commercial versions usage text resources from owners who have granted a consortium license, as well as resources that, in accordance with the Copyright and Related Rights Act and EU regulations, can be utilized to build a full open model.
The PLLuM technological models (i.e. available on licenses that do not let commercial applications) besides usage publically available data sets specified as Common Crawl.
According to the digitisation department, PLLuM together with the model Whitehead can advance artificial intelligence created in Poland, supporting each another in a better process of training and further obtaining and beginning data needed to #AIMadeInPoland was improving – for public administration, business and society.
Here is Bielik – Poles besides have their own AI based on LLM
– improvement of PLLuM is an investment in a digital state. So far we have allocated PLN 14.5 million for this project, and now we are going a step further – another PLN 19 million will let us to implement the model in public administration and extend cooperation with fresh partners specified as COI and Digital. This will make PLLuM a key component in the digitisation of public services and the improvement of the national AI ecosystem," said Deputy Minister of Digitalization Dariusz Standerski.
The task is implemented on behalf of the Ministry of Digital Affairs, which owns the results and controls the improvement of PLLuM. The task has so far been implemented by a consortium of six entities:
- Wrocław University of Technology (project leader)
- Institute of Computer discipline PAS
- Institute of Slavics PAS
- Scientific and Academic Computer Network (NASK-PIB)
- Information Processing Centre (OPI-PIB)
- University of Lodz
Where will the PLLuM go? 1 of the ideas is the virtual assistant function built into the future application iterations mCitizen, which is designed to make it easier for users to access public information. The Ministry besides sees the application of the ready-to-use household of Polish AI models in the administrative or education sectors.
PLLuM is available at: http://pllum.clarin-pl.eu. Models can be downloaded on Hugging Face.
If article The Ministry of Digital Affairs announces PLLuM – a household of Polish artificial intelligence models – can already be used does not look right in your RSS reader, then see it on iMagazine.
