News

On the occasion of Meltemi, the first Large Language Model for the Greek language created by researchers at the Athena Research Center, the Director of the Institute for Language and Speech Processing at Athena Research Center, Vassilis Katsouros, gave an engaging interview to Naftemporiki TV and journalist Takis Spiliopoulos on Monday, June 17, 2024.

 

Meltemi is the first Greek system equivalent to chatGPT. As Mr. Katsouros explained, it has been fed with Greek texts comprising billions of words from a vast array of sources and is being trained and improved daily. GPTs (Generative Pre-trained Transformers) are generative models structured with neural network architecture consisting of billions of parameters, requiring large volumes of data and texts for their training. However, Mr. Katsouros emphasized the importance of utilizing quality data for training the model.

 

 

Currently, in Europe, there is a significant effort to create a multilingual Large Foundational Language Model supporting European languages. In this endeavor, the Athena Research Center collaborates closely with the Ministry of Digital Governance for the European Language Alliance. As the ILSP Director highlighted, Greece also needs to form an alliance for the Greek language.

 

Looking ahead, Mr. Katsouros mentioned that the goal is for Large Language Models to become multimodal, meaning they would be capable of understanding images, shapes, mathematical equations, sound, video, and music.

 

Language, as a natural means of human communication and interaction, clearly plays a crucial role in the development of complex artificial intelligence systems.