В Казахстане разработали казахскую языковую модель ИИ Kaz LLM - 24.kz
В Казахстане разработали казахскую языковую модель ИИ Kaz LLM - 24.kz.
- 01.В Казахстане разработали казахскую языковую модель ИИ Kaz LLM - 24.kz.
The President instructed the President to speed up the development of artificial intelligence in our country at an extended meeting of the Cabinet of Ministers. By the end of the year, its elements will be integrated into the e-government system. In Kazakhstan, meanwhile, the Kazakh language model Kaz LLM was developed. This is artificial intelligence that understands and speaks Kazakh. A 24KZ correspondent found out how it differs from foreign analogues. Guldana Kalen, correspondent: - At first glance, the language model is just a tool for working with text, an analogue of the well-known GPT chat, which can process requests, generate responses and solve many other problems. But the main difference between Kaz LLM is that the platform understands all the subtleties of our mentality. That is, she was taught to understand the history and culture of Kazakhstan. All domestic literature, scientific works and modern content from open sources were uploaded to the platform. And all this, according to the developers, made it possible to adapt the model to national characteristics. Madina Abdurakhmanova, senior data analyst at the Institute of Smart Systems and Artificial Intelligence: - Unfortunately, the Kazakh language is one of the low-resource types of languages on the global Internet. That
There is much less information in written sources in the Kazakh language compared to, say, English, French, German or Russian. In order for Kaz LLM to identify our worldview in Kazakhstan - the way we see history, culture, family values, as well as the rights of our fellow citizens - we had to teach her a different vision of the world and a different way of working with the Kazakh language. The language model currently supports four languages: Kazakh, Russian, Turkish and English. She easily translates texts between them. The main advantage of Kaz LLM, according to the developers, is security and data protection. Information will not leave the country's borders. This is especially important for use in government agencies and large corporations, where data leakage can lead to serious risks. Madina Abdurakhmanova, senior data analyst at the Institute of Smart Systems and Artificial Intelligence: - Plus, we live in a time when, unfortunately, there are various sanctions. We have a multi-vector policy, but nevertheless, none of us is immune from the fact that certain services will be suspended simply because of problems with the Internet or because of certain decisions. Therefore, creating a model is artificial
intellectual intelligence within a country is truly a matter of national sovereignty. We must be self-sufficient, we must have our own products, which are made on a completely Kazakh core. The model is now being tested in practice. On its basis, voice assistants are created for government portals, tools for translation and text analysis, as well as document automation systems. According to forecasts of the relevant ministry, in the future Kaz LLM will become the basis for creating products in a variety of fields: from medicine to education. Darkhan Myrzabaev, Deputy Chairman of the Committee for Artificial Intelligence and Innovation Development of the MCRIAP RK: - That is, a person, for example, who wants to create some kind of product in the field of medicine, for example, prescriptions, he needs a chatbot or an assistant for issuing prescriptions, help in how to take medications. In this direction, a person can further train his model using this data, add certain data on pharmaceuticals and create his own product in this direction. To create the model, it took not only modern technologies, but also a team of professionals. The project involved exclusively Kazakh programmers, analyst
ics and researchers. The main work took 8 months, but the base began to be prepared five years ago. Authors: Guldana Kalen, Order of Zharkyn.
Related Intelligence
All stories →«用AI玩具欺骗总统》:阿别诺夫批评哈萨克斯坦官员
哈勒克委员会成员穆拉特·阿别诺夫批评部委和州政府未加载国家超级计算机。据他称,官员们继续使用俄罗斯服务。这是因为他们担心腐败计划会被揭露……
ChatGPT 新增受 PIN 码和 Face ID 保护的加密聊天
OpenAI 正在向 ChatGPT 推出加密聊天功能,允许用户用 PIN 码或生物识别保护隐藏单独的对话。这一新功能有助于保护个人对话不被他人查看,即使用户将手机交给他人也无法访问。
微软的 MAI-DxO 系统诊断复杂疾病比医生更准确
微软开发的 MAI-DxO 人工智能系统在诊断复杂临床病例时表现优于全科医生。内部测试中,该系统的诊断准确率超过 80%,在某些指标上甚至是医生结果的两倍……