If we do not have control over the sources of knowledge that we choose, then someone else will program our young generation – says in the new episode of the “Techpospolita” podcast on TVN24+ dr inż. Jan Kocoń, co -founder of the first Polish language model of the flum. Recently, all Poles can use it, and the importance of the creation of this model is huge.
The first Polish language model has been operating since February. It is based on original Polish data, built on one of the largest text corps in the country and adapted to the specifics of the Polish language and public administration terminology. Why is having such technology in Poland important? Co -creator of the flum, dr inż. Jan Kocoń explains this in the podcast “Techpospolita” on TVN24+: Every country in the world will have artificial intelligence – its own or foreign.
Why is Polish artificial intelligence so important?
In an interview with Tomasz Kanik from TVN24 Dr. Kocoń, an assistant professor at the Department of Artificial Intelligence of the Wrocław University of Technology, drew attention to the importance of language for the action of large language models such as plum. – We have observed in our research that we can significantly improve the quality of the model relative to a specific language if we provide it with enough data in this particular language at the end of training. It was an idea behind the flum – he noted.
This, in turn, is of great importance for deciding the results that a given AI model will show us. The creators of language models are able to tune them so as to show them what they should say and what they should not. Nobody wants AI to give recipes for drug production or the construction of a bomb. But what the language model is to correspond to its creator. In this way, for example, the Chinese Deepseek model does not respond to Questions about the massacre on Tiananmen Square in Beijing.
That is why the creators of the Polish language model have just devoted a lot of attention to training in terms of safety. – (trained) model is starting to understand better (…) how it should give these answers to provide our security policy, and on the other hand to have a certain style that we impose because of our preferences – he explained.
The key in this context is to teach AI, what sources of information he should use. Dr. Jan Kocoń emphasizes how important it is. – Since we acquire knowledge from various sources, and we will not have control over the sources that we will choose, someone else will “program” our young generation – he says directly. Regardless of whether it is “programming young” will take place accidentally or in a purposeful way. – (Therefore,) it's good to have this technology at home, have full control over it – he emphasizes.
Does the flume have views?
When asked if the flum would have any views or whether he would be religious, his co -creator emphasizes that not. – We did not deal with this topic at all (…). We left it not so much apolitical, but we did not interfere in this process related to preferences. This means that depending on how we ask him, we can get such information from it – he admits in an interview with TVN24+. – Of course, I am talking about ideological issues, where it is difficult to find a source of truth, because it is a matter of faith, views – he adds.
The whole conversation – episode of the “Techpospolita” podcast entitled “Artificial intelligence in national colors” – can be viewed on TVN24+.
Author/author: pb // mm
Source: tvn24.pl
Source of the main photo: Adobe Stock