Türkiye has developed its homegrown option in the field of natural language processing (NLP), an artificial intelligence model that has gained attention worldwide and is due to launch its first AI-based chatbot.
The explosive success of generative AI chatbots similar to ChatGPT spurred a global artificial intelligence revolution in late 2022. But it also warned that such tools could infect the internet with disinformation.
Türkiye's Havelsan, known for its software-based tech products in the defense industry, has built a "national GPT" and is planning its launch in the coming days, Information and Communication Technology product engineer Osman Kavaf said on Tuesday.
AI breakthroughs are currently being led by the United States, home to ChatGPT-maker OpenAI, with China close behind.
Türkiye does not want to be left behind in the digital drive, and officials have repeatedly voiced ambitions to grow the country's own AI champions.
ChatGPT has captured world attention with its advanced capabilities, catapulting the San Francisco-based startup to the top ranks of AI companies. Generative AI systems like ChatGPT can spit out new text, images, videos or audio recordings based on prompts from users.
Initiated a year ago, Türkiye's project in NLP, a branch of artificial intelligence that enables computers to understand text and spoken words, focused on GPT models and considered the needs of public and military institutions to ensure data security, Kavaf said.
"Our domestic and national Turkish GPT is ready after a yearlong project. We plan to launch our first version on Feb. 1," he told Anadolu Agency (AA).
Havelsan has been involved in various AI projects, focusing on algorithms for texts, images and sounds, said Kavaf.
Named "MAIN," an acronym derived from the first letters of Multifunctional Artificial Intelligence Network, the product falls under the category of "productive AI."
Havelsan has spearheaded software technology in Türkiye's defense industry for years, which has undergone a profound transformation since the early 2000s.
Over the last two decades, the drive has aimed to reduce external dependency on Western products through innovative engineering initiatives and domestically developed technologies.
Kavaf emphasized that Havelsan's GPT possesses the fundamental features of all existing models, highlighting summarization as one of its key functionalities.
"It can summarize content and text within seconds. It can retrieve information from open-source data and can write code," he stated. "By adjusting the parameters or making changes to your commands, you can diversify these basic functions."
Kavaf underscored that public and military institutions are prioritized on their list, indicating that the application will not be publicly available initially.
"In the subsequent versions of our project, initiated to ensure the data security of public and military institutions, we hope to develop a mechanism that can be publicly accessible," he said.
Regarding the security aspects of the project, Kavaf said in contrast to their global rivals, along with the hardware set to be provided as a device, the company will "ensure that this data is used as a completely closed box."
"This way, we will prevent data leakage or the use of interpreted data anywhere. Currently, there is no access to such a product in the market," he noted.
Kavaf said they decided that the product should progress as a flagship. "While we can integrate the cloud system in subsequent models, it does not solve the issue of using sensitive and highly confidential data.
"We want to stand out with a hardware set, especially in using sensitive and highly confidential data."
The launch date has been strategically set to strengthen the hardware infrastructure, according to Kavaf.
"After the launch, we will rapidly evaluate the demands and needs. On Feb. 1, Havelsan is hosting a major AI conference. It will be a special event where we plan to launch our product," he noted.
Kavaf emphasized that AI would become prevalent in various aspects of life as the initial versions focused mainly on texts and writings.
"Actually, we are preparing the infrastructure for this. There is incredible awareness globally. In Türkiye, this awareness has also risen rapidly," he stated.
"This will be our first version, and in this initial release, we primarily focus on texts and writings. In the subsequent stages, along with image and sound processing features, we will be able to use artificial intelligence more effectively."
