Aleksa Gordic, founder of Runa AI – a multilingual startup co-founder of generative artificial intelligence core models for enterprise – announced the availability of YugoGPT.
Gordic claims that this is the largest productive language model built to teach Serbian, Croatian, Bosnian, and Montenegrin languages to do more or less the same thing.
ChatGPT is for English, meaning to understand texts, answer questions, and act as an AI assistant for people and companies in the region.
Bosnian, Croatian, Montenegrin, and Serbian (BCMS) are mutually intelligible South Slavic languages spoken in Southeastern Europe, namely Bosnia and Herzegovina, Croatia, Montenegro, and Serbia.
“ChatGPT will respond to your inquiry on how you plan to pay your taxes this year, but it will assume that you do so in the United States.
On the other hand, you can train a broad language model (LLM) for your own local needs. This is the advantage of yugoGPT”, Gordic told Serbian media When asked in an interview with Biznis.rs why the locals need it, chatGPT is in their language.
Motivated by a sense of frustration with apparent shortcomings outside of the English Natural Language Processing (NLP) realm, Gordic embarked on the creation of YugoGPT earlier this summer to raise the standard of language models beyond the confines of the English-dominated NLP landscape.
“We believe AI should serve every language, and we build multilingual GenAI/underlying models for businesses,” Aleksa Gordic said in a LinkedIn post.
“Our next step will be to raise seed funding to accelerate, acquire a GPU cluster, and build an enterprise LLM platform,” he added.
Yugo’s Unique AbilitiesGPT
“YugoGPT 7B significantly beats Mistral and ranks #2 from LLaMA Meta (formerly Facebook) and is now officially the world’s best open source LLM for Serbian and other HBS (Croatian, Bosnian, Montenegrin) languages.” si.” Mentioned on LinkedIn.
According to Gordic, model parameters will be made accessible to individuals and companies.
This will increase flexibility by enabling customization for specializations in various fields such as finance, taxes, psychology, and more.
Amid concerns about data security and privacy, numerous companies are expressing hesitation about trusting American APIs, citing discomfort with sending sensitive data to third-party servers.
Gordic sheds light on this trend, noting a growing preference for using AI capabilities directly in on-premises computing systems.
Gordic emphasizes that access to model parameters gives organizations greater control over the functions of AI and addresses concerns about third-party involvement.
Gordic, who graduated from the electronics department of the Belgrade Faculty of Electrical Engineering in 2017, worked as a software and machine learning engineer on the HoloLens project at Microsoft’s Development Center in Serbia
After this, he specialized in languages and joined Google’s DeepMind. Models with image and video understanding capabilities.
